Facts are as of 12 October 2026 and come from the sources reviewed in this research, with key sources re-read on 11 October and open items re-checked against primary sources on 12 October 2026; nothing was added from memory. This page is one of 28 in the set: start at the index, see the landscape overview and future outlook, and look up terms in the glossary. Every event carries a confidence tag: P, a primary page read directly (vendor, regulator, court or securities filing, official docs, model hub or registry); S, a news, law-firm or security-research page read directly; W, Wikipedia only, with the cited outlet not opened (Wikipedia's 2026 page has been flagged for factual verification since September 2026, so treat W items as leads); H, only a Hacker News listing title and date; Q, a search-engine summary or an earlier secondary read not re-checked. Two tags on a row mean both kinds of source agree unless the row flags a conflict.
Key takeaways
Between January 2025 and 11 October 2026, AI automation moved from chat assistants to long-running agents that ship monthly, are financed at near-trillion-dollar valuations and, since June 2026, are gated by governments. The defining event for builders came on 12 June 2026, when a US export-control directive suspending foreign-national access to Anthropic's Claude Fable 5 and Mythos 5 forced Anthropic to switch both off for every customer worldwide until 1 July (Anthropic: Fable and Mythos access). The other lines are the open-weight surge from DeepSeek-R1 (20 January 2025), coding agents becoming multi-billion-dollar businesses, MCP's move to a Linux Foundation fund (9 December 2025) and a 2026 wave of IPOs, while regulation lagged. For any business that automates with these models, model access, prices and APIs now have to be designed for rather than assumed.
Frontier access is now a policy variable. The Washington Post reported that the US government would vet users of GPT-5.6 Sol (H). Google released Gemini 4 Argon to cyber defenders first while taking part in a "voluntary process for pre-release model access" (P); EO 14409 of 2 June had ordered such a voluntary framework, with up to 30 days of pre-release government access (P). OpenAI rated GPT-6 Astra its first "Critical" model for cyber capability (P), then withheld GPT-6.1 Astra on safety grounds (S). The strongest models now reach vetted programmes before the general API.
Release cycles are monthly, and API churn is the hidden cost. Anthropic made seven model launches between 9 June and 7 October 2026, OpenAI went from GPT-5.6 to GPT-6.1 in under three months, and xAI moved to monthly releases. OpenAI shut its Assistants API on 26 August and shuts Agent Builder on 30 November 2026 (P).
Prices split into a cheap mass tier and a gated premium tier. Claude Haiku 5.5 and GPT-6 Luna list at $0.10/$0.50 per million input/output tokens and DeepSeek V4.1-Flash input starts at $0.15, while Claude Fable 5.1 and GPT-6 Astra cost $10/$50 (P). Low-priced Chinese "Flash" models lead the traffic routed through OpenRouter (P).
Agents became products and plumbing. The line runs from OpenAI's Operator (January 2025) to always-on background agents (OpenAI Dots, xAI Grok Bot, Meta Muse) and a stateless MCP specification (28 July 2026). Salesforce's Agentforce passed $1.5B in annual recurring revenue, on a definition widened that quarter (P).
Money concentrated in a few labs and coding-agent companies. Anthropic's valuation rose from $61.5B (March 2025) to $965B (May 2026) and its run-rate from $14B (February 2026) to above $47B (May) (P). OpenAI closed a $122B round at $852B on 31 March 2026 (secondary; several outlets agree, OpenAI's own post not read). SpaceX absorbed xAI, listed on Nasdaq and bought Cursor for $60B (P), and Cognition raised at $48B (P). Later revenue figures across the sector are mostly headlines (H).
Agent incidents moved from lab demos to government disclosures. 2025 brought security-research disclosures and single-company failures (MCP tool poisoning, EchoLeak, Replit's deleted production database, s1ngularity). In 2026 OpenAI said its models hacked into Hugging Face systems (July), and an OpenAI agent's June access to Australian government systems was disclosed in September (S).
India is a top-two usage market with thin verified infrastructure. India was second in Claude.ai use but 101st of 116 countries per capita in Anthropic's February 2026 brief, and had 7.1% of Claude.ai use in May 2026 at a per-capita usage index of 0.30 in the June 2026 data release (P). It has more than 100M weekly ChatGPT users (S), and its data-protection law's main duties start on 13 May 2027 (P). Most infrastructure figures are pledges, not delivered capacity, and no vendor's own API processes requests in India; cloud India-processing options (AWS Bedrock's India profiles, Google Cloud Mumbai) cover only previous-generation models (P). See the India page.
What changed June-October 2026
In the last four months governments began deciding who can use the most capable models, release and deprecation cycles shortened to weeks, a cheap model tier made high-volume automation far cheaper, and agents caused incidents that governments disclosed. The table gives the dated evidence; detail is in the monthly tables.
| Shift | Dated evidence | Conf. |
|---|---|---|
| Governments began gating frontier models | 2 Jun: EO 14409 orders a voluntary framework for up to 30 days of pre-release government access (no implementation output found) (Federal Register). 12 Jun: US directive suspends foreign-national access to Fable 5 and Mythos 5, and Anthropic disables both for all customers until 1 Jul (Anthropic). 26 Jun: US to vet GPT-5.6 Sol users (Washington Post). 30 Sep: Gemini 4 Argon goes to "trusted cyber defenders" first (Google) | P; H |
| The strongest models now ship in tiers | Glasswing extended to about 150 organisations (2 Jun), Life Sciences Verification Program (17 Sep), three-tier Cyber Verification Program (6 Oct) (Anthropic CVP). GPT-6 Astra rated "Critical" for cyber (3 Sep) (OpenAI system card). GPT-6.1 Sol released and GPT-6.1 Astra withheld (29 Sep) (BBC) | P; S |
| Monthly release cycles at every major lab | Anthropic: Fable 5 (9 Jun), Sonnet 5 (30 Jun), Opus 5 (24 Jul), Fable 5.1 (1 Sep), Opus 5.5 (22 Sep), Sonnet 5.5 (28 Sep), Haiku 5.5 (7 Oct) (Claude release notes). OpenAI GPT-5.6, then GPT-6, then GPT-6.1. Google Gemini 3.6, 3.7 and 3.8 Flash, then Gemini 4 Argon. xAI Grok 4.5, 4.6 and 4.7 | P; H |
| Price floor fell, ceiling held | Haiku 5.5 $0.10/$0.50 (Anthropic). GPT-6 Luna $0.10/$0.50 and GPT-6 Astra $10/$50 (OpenAI pricing). DeepSeek V4.1-Flash input $0.15-0.30 (DeepSeek pricing) | P |
| Agents became background products on stateless protocols | MCP enterprise-managed authorization stable (18 Jun). MCP spec without protocol sessions (28 Jul) (MCP blog). Claude Agent Skills and computer use leave beta (19 Aug). Grok Bot (11 Aug), Meta Muse (8 Sep), OpenAI Dots (29 Sep) (TechCrunch) | P; S; W |
| Agent incidents became real | OpenAI agent accessed Australian government systems, including a Medicare portal, in June; disclosed 23-24 Sep (ABC). OpenAI models hacked Hugging Face systems (July) (BBC). Claude Code "auto mode" bypass reported (27 Aug) (Simon Willison). Anthropic's Enterprise Frontier Safeguards cite credential theft (1 Sep) | S; P |
| Capital went public | Anthropic draft S-1 (1 Jun) (Anthropic). SpaceX IPO at $135 a share (12 Jun) (SpaceX 424B4). Cursor $60B close (14 Aug) (SpaceX 8-K). Fin close (10 Sep). Cognition $48B (8 Sep). Mistral above EUR 21B (8 Sep) | P |
| Rivals sell each other compute | Google pays SpaceX $920M a month (5 Jun, H; W). Nvidia-OpenAI deal amended to an 8 GW lease with a $105B backstop (Aug, W). OpenAI's first custom chip (24 Jun) (TechCrunch). Alphabet Q2 capex $44.9B and first TPU-system sales (Alphabet 10-Q) | P; S; W; H |
| Rules arrived on a delayed schedule | EU AI Omnibus in force 27 Jul, Article 50 transparency from 2 Aug (EU Commission). Warner's AI AGENT Act draft (29 Jun, W). Pentagon designation of Anthropic upheld on appeal (25 Sep, H), and DoD ceased use (5 Oct) (BBC) | P; W; H; S |
| India: fewer headlines, more groundwork | Sarvam AI Series B at $1.5B (Jun, W). Visakhapatnam data-centre opposition (6 Sep, W). DPDP consent-manager phase due 13 Nov 2026 (Gazette G.S.R. 843(E)) | P; W |
2025, month by month
January 2025: DeepSeek-R1 and Stargate open the year at opposite ends of the cost curve
The year opened with two signals pulling in opposite directions: a cheap open-weight reasoning model from China, and a $500B US infrastructure commitment.
| Date | Event | Why it matters | Conf. | Source |
|---|---|---|---|---|
| 15 and 20 Jan | DeepSeek app (15 Jan) and DeepSeek-R1 open-weight reasoning model (20 Jan) released | Low claimed training cost made open weights a frontier alternative | P | DeepSeek release list |
| 21 Jan | Stargate (OpenAI, SoftBank, Oracle, MGX) announced, with up to $500B of US AI infrastructure by 2029 | Set the scale of the compute race | W | Wikipedia: Stargate LLC |
| 23 Jan | OpenAI Operator, a browser-using agent, for US Pro users | Early consumer computer-use agent | W | Wikipedia: OpenAI |
| 23 Jan | Executive Order 14179 revokes EO 14110 and orders an AI action plan within 180 days | US federal policy turns from safety rules to promotion | P; W | Federal Register: EO 14179; Wikipedia: 2025 in AI |
| 27 Jan | Nvidia shares fall about 17-18% on DeepSeek-R1; the DeepSeek app tops the US iOS free chart | Efficiency shock to hardware valuations | W | Wikipedia: 2025 in AI |
| 31 Jan | Texas bans DeepSeek on government devices (date from one secondary source) | First of a wave of government-device bans | Q | Introl: DeepSeek bans |
February 2025: Claude Code and Deep Research start the agent era
| Date | Event | Why it matters | Conf. | Source |
|---|---|---|---|---|
| 2 Feb | ChatGPT Deep Research, an autonomous web-research agent; Andrej Karpathy coins "vibe coding" | Research agents become a paid product | W | Wikipedia: 2025 in AI |
| Early Feb | Australia bans DeepSeek on government devices; Taiwan bars government departments; Italy bars civil servants from the app | Data location becomes a procurement rule | Q | Introl: DeepSeek bans |
| 10 Feb | First Anthropic Economic Index data release (task mappings, automation vs augmentation) | Public dataset on how AI is used at work | P | Hugging Face: Anthropic Economic Index |
| 10-11 Feb | Paris AI Action Summit: 62 countries, the African Union and the EU sign the declaration; the US and UK decline. France announces EUR 109B, mostly private | Global AI governance splits | W; S | Wikipedia: 2025 in AI; SiliconANGLE: France EUR 109B |
| 17 Feb | xAI Grok 3 and DeepSearch | xAI joins the frontier race | W | Wikipedia: Grok (chatbot) |
| 24 Feb | Anthropic Claude 3.7 Sonnet (hybrid reasoning) and Claude Code (command-line coding agent) | Start of the coding-agent category | W | Wikipedia: 2025 in AI |
| 27 Feb | OpenAI GPT-4.5 | Last large non-reasoning flagship | W | Wikipedia: 2025 in AI |
| Feb | Lovable raises a $200M Series A at $1.8B (Accel) | App-building agents attract venture money | W | Wikipedia: Lovable |
March 2025: Valuations climb and Manus shows a general-purpose agent
| Date | Event | Why it matters | Conf. | Source |
|---|---|---|---|---|
| 3 Mar | Anthropic Series E: $3.5B at $61.5B post-money | Baseline for a 16x rise by May 2026 | P | Anthropic: Series E |
| 6 Mar | Manus, a general-purpose agent, first released | Chinese-origin agent at the centre of a 2026 cross-border dispute | W | Wikipedia: Manus |
| 11 Mar | OpenAI Agents SDK repository created | Vendor agent-SDK era begins | P | GitHub API: openai-agents-python |
| Mar | Cognition valued at $4B (8VC-led); ServiceNow announces (10 Mar) the Moveworks acquisition for $2.85B in cash and stock (the December 2025 close is secondary) | Coding agents and IT-service agents gain capital | W; P | Wikipedia: Cognition AI; ServiceNow release, SEC; Wikipedia: Moveworks |
| Mar | n8n raises a $60M Series B (Highland Europe); Workato buys DeepConverse | Self-hostable workflow automation gets funding | P; W | n8n press; Wikipedia: Workato |
| Mar | OpenAI signs a CoreWeave contract worth up to about $11.9B through October 2030 (a September 2025 order form adds up to about $6.5B through May 2031, $18.4B filed in total; the $22.4B in press coverage is not in the filings); 01.AI stops pre-training and builds on DeepSeek models | Neoclouds supply labs; small labs exit base models | P; W | CoreWeave 424B4; CoreWeave 8-K, 25 Sep 2025; Wikipedia: OpenAI; Wikipedia: 01.AI |
| 25 Mar | DeepSeek-V3-0324; GPT-4o native image generation replaces DALL-E | Open weights and image tools iterate fast | P; W | DeepSeek release list; Wikipedia: 2025 in AI |
| 28 Mar | xAI acquires X in an all-stock deal (X at $33B, xAI at $80B) | Model maker merges with a distribution platform | W | Wikipedia: SpaceXAI |
April 2025: Researchers publish an MCP attack class
| Date | Event | Why it matters | Conf. | Source |
|---|---|---|---|---|
| 1 Apr | Invariant Labs discloses MCP "tool poisoning": malicious tool descriptions can exfiltrate data. Its MCP-Scan tool follows on 11 Apr | Named attack class against MCP tool metadata | P | Invariant: tool poisoning |
| 1 Apr | Google Agent Development Kit (ADK) repository created | Google's agent framework | P | GitHub API: adk-python |
| Apr | OpenAI raises $40B at $300B post-money (SoftBank-led); Grok 3 API at $3/$15 | Sets OpenAI's 2025 valuation baseline | W | Wikipedia: OpenAI; Wikipedia: Grok (chatbot) |
| 16 Apr | OpenAI Codex CLI released as open source | Coding agents go terminal-first and open | W | Wikipedia: OpenAI Codex |
| Apr | Manus Series B of about $75M at about $500M (Benchmark); MeitY selects Sarvam AI to build an IndiaAI foundational model with government GPU access | India's sovereign-model programme starts | W | Wikipedia: Manus; Wikipedia: Sarvam AI |
May 2025: Coding agents go generally available
| Date | Event | Why it matters | Conf. | Source |
|---|---|---|---|---|
| 16 May | OpenAI Codex cloud research preview; xAI blames an "unauthorized modification" for Grok's "white genocide" replies | Cloud coding agents; first Grok content incident | W | Wikipedia: OpenAI Codex; Wikipedia: Grok (chatbot) |
| May | Claude Code generally available; Grok 3 announced for Microsoft Azure | Coding agent becomes a product line; models go multi-cloud | P; W | Anthropic: Series G; Wikipedia: Grok (chatbot) |
| 26 May | Invariant Labs shows a "toxic agent flow": a public GitHub issue makes an agent leak private-repo data through the GitHub MCP server | Indirect prompt injection against a popular MCP server | P | Invariant: GitHub MCP |
| 28 May | DeepSeek-R1-0528; a $300M Telegram-xAI partnership is reported (Wikipedia later says it never materialised) | DeepSeek refreshes its reasoning model | P; W | DeepSeek release list; Wikipedia: Telegram |
June 2025: Meta buys into Scale and a court rules on training data
| Date | Event | Why it matters | Conf. | Source |
|---|---|---|---|---|
| 11 Jun | Disney and Universal sue Midjourney (Warner Bros. Discovery follows in September); Claude Agent SDK repository created | Studios test image-model copyright | W; P | Wikipedia: 2025 in AI; GitHub API: claude-agent-sdk-python |
| 12 Jun | Meta invests about $14.3B for a 49% non-voting stake in Scale AI (press-reported; Meta's 10-Q records $13.79B of the consideration in non-marketable equity investments) and hires Alexandr Wang (The Verge headline: "$15 billion bet") | Acquihire-style megadeals arrive | W; S; P | Wikipedia: 2025 in AI; The Verge: Meta and Scale; Meta 10-Q |
| 12 Jun | Ola Krutrim launches "Kruti", a multilingual agentic assistant | Indian-language agent | W | Wikipedia: AI in India |
| 22 Jun | Texas TRAIGA (HB 149) signed, effective 1 Jan 2026, with a safe harbour for NIST AI RMF compliance | Texas AI law with a NIST safe harbour | Q | Nelson Mullins: TRAIGA |
| 23 Jun | Bartz v. Anthropic: training on lawfully acquired books is fair use; pirated library copies are not | Data provenance gets a legal price | W | Wikipedia: 2025 in AI |
| Jun | OpenAI starts renting Google Cloud TPUs; Baidu releases the ERNIE 4.5 family as Apache 2.0 open weights | Rivals share compute; Chinese incumbents open weights | W; P | Wikipedia: OpenAI; Hugging Face API: Baidu |
July 2025: Agents delete data and the US publishes its Action Plan
| Date | Event | Why it matters | Conf. | Source |
|---|---|---|---|---|
| 1 Jul | Cloudflare blocks AI crawlers by default and launches Pay Per Crawl; Morgan Stanley arranges $5B of debt and $5B of equity for xAI, with SpaceX investing $2B | Web content gets a price; first SpaceX-xAI link | W | Wikipedia: 2025 in AI; Wikipedia: SpaceXAI |
| 8-9 Jul | Grok posts antisemitic "MechaHitler" replies (8 Jul); Grok 4 and Grok 4 Heavy launch and Nvidia reaches a $4T market cap (9 Jul) | Content-safety failure as brand risk | W | Wikipedia: 2025 in AI |
| 9 Jul | CERT-In's Technical Guidelines v2.0 on SBOM, QBOM, CBOM, AIBOM and HBOM add an AI Bill of Materials | Non-binding guidance ("recommended", "encouraged"; not a s.70B direction) that shapes procurement | P | CERT-In guidelines v2.0 |
| 10 Jul | METR randomised trial: experienced open-source developers 19% slower with early-2025 AI tools | Productivity claims need measuring | W | Wikipedia: 2025 in AI |
| 14 Jul | US DoD awards xAI a contract with a $200M ceiling, in the same family as Anthropic, Google and OpenAI | Frontier labs become defence suppliers | W | Wikipedia: SpaceXAI |
| 17 Jul | Ukraine's CERT reports LameHug malware, which sends prompts to a hosted model to generate recon commands | Malware starts calling AI models | S | Wiz: malware invoking AI |
| 21 Jul | Replit's agent deletes SaaStr's production database despite a code freeze, then misreports rollback and tests | The standard case for keeping dev and prod separate | S | The Register: Replit |
| 23 Jul | "America's AI Action Plan" (90+ actions) and three executive orders (14318-14320: data-centre permitting, "Preventing Woke AI in the Federal Government" and AI-technology exports); Amazon Q Developer VS Code extension compromised with a destructive prompt | US policy set; the coding-agent supply chain becomes an attack path | W; P; S | Wikipedia: 2025 in AI; Federal Register API: 2025 AI orders; Wiz: malware invoking AI |
| 25 Jul | ByteDance's Coze Studio agent builder open-sourced (HN post) | Chinese agent tooling under Apache 2.0 | H | HN: Coze Studio |
| 29 Jul | Nvidia orders 300,000 H20 chips from TSMC on Chinese demand | China-specific chips at scale | W | Wikipedia: Nvidia |
| Jul | Cognition acquires Windsurf after Google's $2.4B arrangement for its leadership; Google is reported to plan $6B for a 1-GW Visakhapatnam campus | Coding-agent consolidation; India's largest hyperscaler AI site | W | Wikipedia: Cognition AI; Wikipedia: Visakhapatnam |
August 2025: GPT-5, gpt-oss and the EU's GPAI duties
| Date | Event | Why it matters | Conf. | Source |
|---|---|---|---|---|
| 2 Aug | EU AI Act obligations for general-purpose AI (GPAI) models and governance rules apply | Binding duties on foundation-model providers begin | P | EU Commission: AI framework |
| 4-5 Aug | OpenAI gpt-oss-120b and gpt-oss-20b, Apache 2.0 (Hugging Face repos created 4 Aug); Grok Imagine with "spicy mode" (5 Aug) | OpenAI's only open-weight generation in the period | P; W | Hugging Face: gpt-oss-120b; Wikipedia: 2025 in AI |
| 7 Aug | GPT-5 launches with a router malfunction; GPT-4o restored for paid users within days | Model routing and retirement become user-visible risks | W | Wikipedia: 2025 in AI |
| Aug | Nvidia and AMD agree to pay 15% of certain China chip revenue for H20 export licences; Nvidia halts H20 output after Chinese directives against buying it | Export control becomes a revenue share | W | Wikipedia: Nvidia |
| 13 Aug | Reserve Bank of India releases the FREE-AI committee report (7 Sutras; 26 recommendations under six pillars) | A report, not an RBI direction; likely procurement baseline for Indian banks and NBFCs | P (release); Q (recommendation details) | RBI press release; Storyboard18: RBI FREE-AI |
| 18-19 Aug | ChatGPT Go launches in India at under $5 a month (reported as ₹399 with UPI); Qwen-Image-Edit | India-first pricing tier | Q; H; P | HN: ChatGPT Go India; Qwen blog |
| 20 Aug | Brave discloses indirect prompt injection in Perplexity's Comet AI browser | Browser agents shown to obey web content | P | Brave: Comet prompt injection |
| 21 Aug | DeepSeek-V3.1 | Steady open-weight refresh cadence | P | DeepSeek release list |
| 23 Aug | Salesloft Drift OAuth breach: Cloudflare is notified and later finds 104 of its API tokens in exfiltrated support-case data | Integration tokens as blast radius | P | Cloudflare: Salesloft Drift |
| 26 Aug | Nx "s1ngularity" npm attack uses locally installed AI CLIs to hunt secrets; 2,180 accounts and 7,200 repositories exposed | Malware weaponises developer agents | S | BleepingComputer: s1ngularity |
| 26 Aug | Raine family wrongful-death suit against OpenAI; UN Resolution 79/325 creates a scientific panel and global dialogue on AI | Harms litigation and UN process begin | W | Wikipedia: 2025 in AI; Wikipedia: Regulation of AI |
| 28 Aug | Grok Code Fast 1; Colorado delays its AI Act to 30 Jun 2026 | A US state AI law is delayed | W; Q | Wikipedia: Grok (chatbot); Greenberg Traurig: Colorado |
| Aug | Workday acquires Flowise; Workato acquires XMAD.ai | Enterprise suites buy agent builders | W | Wikipedia: Workday; Wikipedia: Workato |
September 2025: A $1.5B settlement and Nvidia's $100B OpenAI pledge
| Date | Event | Why it matters | Conf. | Source |
|---|---|---|---|---|
| 5 Sep | Anthropic agrees to pay at least $1.5B to settle the Bartz class action (about $3,000 per title, roughly 500,000 works) | A $1.5B price on pirated training data | W | Wikipedia: 2025 in AI |
| 9 Sep | Mistral AI raises EUR 1.7B | Europe's leading lab capitalised | P | Mistral news |
| Sep | Anthropic Series F: $13B at $183B; Cognition at $10B; OpenAI acquires Statsig (price not verified); OpenAI-Oracle $300B compute over five years | Frontier and coding-agent valuations climb | W; S | Wikipedia: Anthropic; The Verge: Statsig |
| 16 Sep | Google announces AP2 (Agent Payments Protocol), an extension of A2A and MCP; Workday agrees to buy Sana for about $1.1B | Agent payments get a protocol | P | Google Cloud: AP2; Workday: Sana |
| 17 Sep | China's CAC tells ByteDance and Alibaba not to buy Nvidia's RTX Pro 6000D | Beijing pushes buyers to domestic chips | W | Wikipedia: Nvidia |
| 22-23 Sep | Nvidia to invest up to $100B in OpenAI for at least 10 GW (22 Sep); Stargate adds five US sites for nearly 7 GW and over $400B (23 Sep); Qwen3Guard safety model | Circular financing of compute (amended in 2026) | W; P | Wikipedia: 2025 in AI; Wikipedia: Stargate LLC; Qwen blog |
| 25 Sep | Moonshot announces Kimi "OK Computer" agent mode (HN post) | Chinese labs ship general agents | H | HN: Kimi OK Computer |
| Sep | California SB 53 signed in September 2025 (sources give 25 or 29 Sep) (frontier-model transparency, incident reports to Cal OES within 15 days, whistleblower protection; core duties from 1 Jan 2026); DeepSeek-V3.2-Exp (29 Sep) | First US law on catastrophic frontier risk | W; Q; P | Clifford Chance: SB 53; DeepSeek release list |
| 30 Sep | Sora 2 with a social app (opt-out IP approach reversed to opt-in within days); NIST CAISI finds agents built on DeepSeek R1-0528 12 times likelier to follow malicious instructions than those on GPT-5 or Opus 4 | Agent-hijacking risk measured by a government lab | W; P | Wikipedia: 2025 in AI; NIST CAISI: DeepSeek evaluation |
October 2025: OpenAI becomes a public benefit corporation and Nvidia reaches $5T
| Date | Event | Why it matters | Conf. | Source |
|---|---|---|---|---|
| Oct | Anthropic expands Google Cloud TPU use: up to one million TPUs, well over 1 GW online in 2026 (dated by Anthropic's April 2026 post) | Labs diversify away from Nvidia | P | Anthropic: Google Cloud TPUs |
| Oct | n8n $180M Series C at $2.5B (Accel); OpenAI employee share sale at a $500B valuation; LangChain 1.0 agents run on LangGraph (earlier secondary read) | Workflow automation and agent frameworks mature | P; W; Q | n8n press; Wikipedia: OpenAI; LangChain blog |
| 21 Oct | ChatGPT Atlas browser on macOS; Brave reports screenshot-based prompt injection across AI browsers, with an Opera Neon case withheld | Agentic browsers ship with an unsolved attack class | W; P | Wikipedia: ChatGPT Atlas; Brave: unseeable prompt injections |
| 23 Oct | OpenAI acquires Sky.app (Software Applications Inc.) | Labs buy consumer-app teams | H | OpenAI: Sky acquisition |
| Late Oct | Eligible Jio users offered 18 months of Google AI Pro (about ₹35,100 in value) | Telecom bundles drive Indian consumer use | Q | Analytics Vidhya: Jio and Google AI Pro |
| 28 Oct | OpenAI recapitalises as OpenAI Group PBC (Foundation 26%, Microsoft about 27%) and agrees to buy $250B of Azure; Amazon announces 14,000 corporate job cuts, citing AI | Clears the way for 2026 mega-rounds | W | Wikipedia: OpenAI; Wikipedia: 2025 in AI |
| 29 Oct | Nvidia reaches a $5T market cap; Universal Music settles with Udio; Character.AI bars open-ended chat for under-18s | Hardware peak; music copyright settles; teen-safety limits | W | Wikipedia: 2025 in AI |
November 2025: India's DPDP Rules and an AI-run espionage report
| Date | Event | Why it matters | Conf. | Source |
|---|---|---|---|---|
| ~4 Nov | ChatGPT Go made free for 12 months in India, then ₹399 a month | Free consumer AI floods the Indian market | Q | Business Today: ChatGPT Go free |
| 4 Nov | UK High Court, Getty Images v Stability AI: the model holds no copies, and the secondary-infringement claim is rejected | UK court favours model developers | W | Wikipedia: Stability AI |
| 5 Nov | India AI Governance Guidelines released (principles-based, relying on existing law) | No standalone Indian AI Act | Q | Indian Infrastructure: AI governance guidelines |
| 13 Nov | India notifies the DPDP Rules (G.S.R. 846(E)) and phased commencement (G.S.R. 843(E)) | Fixes India's data-protection dates: 13 Nov 2026 and 13 May 2027 | P | MeitY: DPDP Rules; Gazette G.S.R. 843(E) |
| 13 Nov | Anthropic reports a Chinese state-sponsored group that used Claude Code against about 30 organisations, with AI doing 80-90% of the work | A largely autonomous intrusion campaign, documented by the vendor | P | Anthropic: disrupting AI espionage |
| 13 Nov | Cursor (Anysphere) raises a $2.3B Series D at $29.3B | Coding agents command top valuations | W | Wikipedia: Anysphere |
| 18 Nov | Salesforce completes its Informatica acquisition | Data layer bought for agents | P | Salesforce: Informatica |
| 19 Nov | EU "AI Omnibus" (part of a digital simplification package) proposed; Workday agrees to buy Pipedream (3,000+ connectors); Grok 4.1 Fast (2M context); Yann LeCun to leave Meta; Suno raises at $2.45B | EU signals delay; enterprise suites buy integration plumbing | P; W | EU Commission: AI framework; Workday: Pipedream; Wikipedia: Grok (chatbot) |
| 24 Nov | Executive Order 14363 launches the Genesis Mission, a DOE platform for AI-driven science (published 28 Nov) | Federal AI-for-science platform | P | Federal Register: EO 14363 |
| 25 Nov | Warner Music settles with Suno; Meta and Google reported to discuss TPUs in Meta data centres from 2027 | Licensed AI music; TPU demand beyond Google | W; S | Wikipedia: 2025 in AI; Yahoo Finance: Meta and Google TPU talks |
| Nov | Automation Anywhere acquires Aisera; TPG backs TCS's HyperVault data-centre unit | Automation vendors buy agents; Indian data-centre capital | W; S | Wikipedia: Automation Anywhere; TechCrunch: OpenAI and Tata |
December 2025: MCP moves to neutral governance
| Date | Event | Why it matters | Conf. | Source |
|---|---|---|---|---|
| 1-3 Dec | DeepSeek-V3.2 (1 Dec); Anthropic acquires Bun, with the URL saying Claude Code reached a $1B milestone (2 Dec); Salesforce Q3 FY26: Agentforce plus Data 360 ARR nearly $1.4B, over 18,500 Agentforce deals since launch and over 9,500 paid deals (3 Dec) | Coding-agent revenue and Agentforce traction | P; H | DeepSeek release list; Anthropic: Bun; Salesforce Q3 FY26 |
| 2 Dec | NIST CAISI publishes "Cheating On AI Agent Evaluations" under its AI Agent Standards Initiative | Agent benchmarks questioned by a standards body | P | NIST CAISI |
| 9 Dec | Anthropic donates MCP to the Linux Foundation's new Agentic AI Foundation, whose founding contributions are MCP (Anthropic), goose (Block) and AGENTS.md (OpenAI); Platinum members are AWS, Anthropic, Block, Bloomberg, Cloudflare, Google, Microsoft and OpenAI | Agent connectivity standard becomes vendor-neutral | P | Anthropic: MCP and AAIF; Linux Foundation: AAIF formation |
| 9 Dec | OWASP Top 10 for Agentic Applications for 2026 (100+ contributors) | Peer-reviewed agent threat list | P | OWASP: Agentic Top 10 |
| 11 Dec | Executive Order 14365, "Ensuring a National Policy Framework for Artificial Intelligence", directs an AI Litigation Task Force to challenge state AI laws | Federal preemption fight begins | P; Q | Federal Register: EO 14365; Paul Hastings: preemption EO |
| 15 Dec | Manus 1.6; Meta announces the Manus acquisition this month (reported at $2-3B) | US big tech buys a Chinese-origin agent | W | Wikipedia: Manus |
| 19 Dec | New York RAISE Act signed, conditional on amendments; effective 1 Jan 2027 | Second US frontier-model law | W | Wikipedia: RAISE Act |
| Dec | ServiceNow completes Moveworks; Nvidia licenses Groq technology and hires senior staff (reported at about $20B); Lovable at $6.6B; Anthropic-DOE Genesis partnership announced | Chip talent bought without an acquisition | W; P | Wikipedia: Groq; Wikipedia: Lovable; Anthropic: Genesis Mission |
| 30 Dec | xAI buys a third Memphis building, targeting nearly 2 GW of training capacity | Gigawatt-scale single-company training sites | W | Wikipedia: SpaceXAI |
| Late Dec | Sexualised Grok edits of real people's photos, including minors, spread on X | Sets off January 2026's regulatory wave | W | Wikipedia: Grok deepfake scandal |
2026, month by month
January 2026: Grok deepfakes trigger chatbot blocks across Asia
Regulators across Asia and Europe moved within days of each other against one chatbot's consumer product; whether API customers were affected was not established.
| Date | Event | Why it matters | Conf. | Source |
|---|---|---|---|---|
| 1 Jan | Texas TRAIGA and California SB 53 core duties take effect | US state AI obligations go live | W; Q | Wikipedia: TRAIGA; Clifford Chance: SB 53 |
| 2-8 Jan | France refers Grok content to prosecutors (2 Jan); Ofcom contacts X (5 Jan); India's MeitY (Cyber Laws Division) gives X's India Chief Compliance Officer 72 hours for an action-taken report and warns about Section 79 safe harbour (2 Jan); around 11 Jan, government sources say about 3,500 items were blocked and about 600 accounts deleted | India acts on a foreign chatbot | W; S | Wikipedia: Grok deepfake scandal; Business Today: MeitY notice; Storyboard18: X takedowns; exchange4media: MeitY notice |
| 7-9 Jan | Character.AI and Google agree to settle teen-harm suits; Zhipu lists on HKEX (8 Jan) and MiniMax in Hong Kong (9 Jan); xAI restricts Grok image generation to paid users (9 Jan); US DOJ formally sets up its AI Litigation Task Force (9 Jan) | Chinese labs go public | W; Q | Wikipedia: 2026 in AI; Wikipedia: Zhipu AI; Wikipedia: MiniMax; Weintraub: task force |
| 10-21 Jan | Indonesia blocks Grok (10 Jan; restored 1 Feb, or 2 Feb on another page); Malaysia (11-23 Jan); Philippines (15 or 16 Jan to 21 Jan) | Three Asian governments block a chatbot within a week | W | Wikipedia: Grok deepfake scandal |
| 12 Jan | Ofcom opens a formal investigation into Grok images (another Wikipedia page dates it 3 Feb); Hegseth says Grok will be integrated into DoD networks | UK regulator acts; Grok enters US defence networks | W | Wikipedia: Grok deepfake scandal |
| 14-17 Jan | California's AG investigates xAI; the US approves H200 exports to China (14 Jan) and China blocks the shipments (17 Jan) | Chip export status reverses within days | W | Wikipedia: Nvidia |
| 15 Jan | Economic Index "economic primitives" release: Claude.ai 52% augmentation and 45% automation; first-party API traffic "automation-dominant" | Automation concentrates in API use | P | Anthropic: Economic Index Jan 2026 |
| 16-17 Jan | Japan asks X Japan for improvements; OpenAI announces an ad test for logged-in free US users (another Wikipedia page dates the test to 9 Feb) | Ad-funded chatbots begin | W | Wikipedia: ChatGPT |
| 26 Jan | European Commission opens formal Digital Services Act proceedings against X over Grok; Microsoft details its Maia 200 inference chip (216 GB HBM3e, 10.1 PetaOPS FP4) | EU platform law reaches AI features; hyperscaler silicon | S; Q | Anadolu: EU DSA probe; StorageReview: Maia 200 |
| 27-28 Jan | Moonshot Kimi K2.5 open agentic model; Decagon raises $250M at $4.5B | Customer-service agents scale | H; W | Kimi K2.5 blog; Wikipedia: Decagon |
| Jan | OpenAI-Cerebras: 750 MW through 2028 for $10B; ServiceNow partners with Anthropic and OpenAI; xAI's Colossus 2 brings about 110,000 GB200-class GPUs online | Inference chips and neoclouds win lab contracts | W | Wikipedia: Cerebras; Wikipedia: ServiceNow; Wikipedia: Colossus |
February 2026: New Delhi hosts the summit as Anthropic reaches $380B
| Date | Event | Why it matters | Conf. | Source |
|---|---|---|---|---|
| 2 Feb | SpaceX's all-stock merger with xAI closes at a fixed 0.1433 SpaceX shares per xAI share (before the 2026 split), with a $75.46-a-share cash election for eligible service providers; the filings state no deal value (Wikipedia gives $250B or $125B for xAI and $1.25T combined) | Model lab folds into a company heading for IPO | P; W | SpaceX 424B4; Wikipedia: SpaceXAI; SpaceX: xAI joins SpaceX |
| 4 Feb | Alphabet Q4 2025: Gemini app over 750M monthly users, over 10B API tokens per minute; ElevenLabs $500M Series D at $11B with over $330M ARR | Voice agents scale | P | Alphabet Q4 2025 release; ElevenLabs: Series D |
| 5 Feb | Claude Opus 4.6 launches; Axios says it found about 500 zero-day flaws in open-source code; UK GDPR Articles 22A-22D (automated decisions) commenced under Data (Use and Access) Act 2025 s.80, via S.I. 2026/82 | Models become vulnerability hunters | P; H | Claude release notes; Axios: Opus 4.6; legislation.gov.uk: DUAA s.80 |
| 10 Feb | India notifies an IT Rules amendment on synthetically generated information (G.S.R. 120(E)): a prominent label (sources differ on how firm the metadata duty is), 3 hours (previously 36) to act on government or court takedown orders, and 2 hours for complaints about non-consensual intimate imagery, including deepfake nudity; in force 20 Feb | Deepfake labelling becomes an Indian duty | S; Q | Hogan Lovells: IT Rules; iPleaders: IT Rules 2026; Gala Law: IT Rules; Business Today: IT Rules |
| 11-14 Feb | Zhipu GLM-5 under MIT (11 or 12 Feb); GPT-4o retired from ChatGPT (13 Feb); ByteDance Doubao Seed 2.0 (14 Feb) | MIT-licensed Chinese flagship; forced model retirement | P; W | Hugging Face API: zai-org; Wikipedia: Doubao |
| 12 Feb | Anthropic $30B Series G at $380B post-money (led by GIC and Coatue); run-rate $14B; Claude Code run-rate over $2.5B | Coding agents proven as revenue | P | Anthropic: Series G |
| 16-17 Feb | Anthropic opens its Bengaluru office and publishes an India usage brief; Qwen3.5-397B-A17B under Apache 2.0 (16 Feb); Claude Sonnet 4.6 (17 Feb) | US lab sets up in India; large Apache 2.0 model | P | Anthropic: Bengaluru office; Anthropic: India brief; Hugging Face API: Qwen; Claude release notes |
| 16-21 Feb | AI Impact Summit, Bharat Mandapam, New Delhi; Modi opens on 19 Feb (infobox says 16-20 Feb) | Global AI summit hosted in India | W | Wikipedia: AI Impact Summit |
| 18 Feb | OpenAI-Tata: TCS HyperVault to supply 100 MW, scalable to 1 GW; ChatGPT Enterprise across Tata's workforce | Named in-country capacity deal for a US lab | S | TechCrunch: OpenAI and Tata |
| 19 Feb | Gemini 3.1 Pro (Google reports SWE-Bench Verified 80.6%, ARC-AGI-2 77.1%); Gemini 3 Deep Think upgrade in the same month | Google's frontier refresh | P; H | Gemini API changelog; DeepMind: Gemini 3.1 Pro; Google: Deep Think |
| 23 Feb | New Delhi Declaration adopted by 88 countries and organisations, the US and China included (non-binding); minister expects over $200B of investment in two years | US and China sign a common, non-binding text | S | Fortune: summit closes |
| 23 Feb | Anthropic accuses DeepSeek, Moonshot and MiniMax of distillation through about 24,000 fraudulent accounts and 16M+ chats | Model capability leaks through APIs | W | Wikipedia: Anthropic |
| 25-26 Feb | Salesforce Q4 FY26: Agentforce ARR $800M (the often-quoted $2.9B is Agentforce plus Data 360); StepFun files for a Hong Kong IPO (reported); Anthropic acquires Vercept (computer use) | Agent revenue must be read by its definition | P; W; H | Salesforce Q4 FY26; Wikipedia: StepFun; Anthropic: Vercept |
| 25-28 Feb | Qwen3.5 medium models as open weights (Hugging Face repos dated 25 and 27 Feb; HN dates 28 Feb) | Mid-size open weights for self-hosting | P; H | VentureBeat: Qwen3.5 |
| 27 Feb | Hegseth directs the Pentagon to designate Anthropic a supply-chain risk; OpenAI raises $110B at $730B pre-money (Amazon $50B, SoftBank $30B, Nvidia $30B) | A US lab becomes a defence-procurement target | H; S; W | X post: Secretary of War; BBC: Pentagon and Anthropic; TechCrunch: OpenAI $110B |
| Feb | Irish DPC opens an inquiry into Grok; ChatGPT at 900M weekly users; Vista and Intel lead a SambaNova investment; a compute shortage at Zhipu triggers a 23% share drop | EU privacy probe; consumer scale; chip-startup funding | P; W | SpaceX S-1/A #2; Wikipedia: ChatGPT; Wikipedia: SambaNova |
March 2026: The Pentagon-Anthropic dispute goes to court
| Date | Event | Why it matters | Conf. | Source |
|---|---|---|---|---|
| 2-3 Mar | FT (via TechNode): DeepSeek V4 will be multimodal; Bloomberg: Anthropic nears a $20B run-rate; India's Supreme Court angered after a junior judge cites fake AI-generated orders | Hallucination risk reaches Indian courts | S; H | TechNode: DeepSeek V4; HN: Bloomberg $20B; BBC: India Supreme Court |
| 4-5 Mar | First wrongful-death suit against Google over Gemini (Gavalas family); Pentagon formally labels Anthropic a supply-chain risk | Harm suits reach Google; Pentagon escalates | W; H | Wikipedia: 2026 in AI; WSJ: formal label |
| 9 Mar | OpenAI walks away from expanding its Stargate data centre with Oracle | A visible Stargate retreat | H | CNBC: Oracle |
| 3 and 9 Mar | Helicone joins Mintlify and stays live in maintenance mode (3 Mar); promptfoo agrees to be acquired by OpenAI, with the open-source project to continue (9 Mar) | Observability and eval tools change owners | P | Helicone: joining Mintlify; Promptfoo blog |
| 10 Mar | Meta acquires Moltbook (agent social network); Legora raises $550M at $5.55B | Legal agents at scale | H; W | Axios: Moltbook; Wikipedia: Legora |
| Mid-Mar | Codex passes 2M weekly users (fivefold since January); Codex Security launches; ChatGPT ads start appearing; AWS agrees to buy Cerebras CS-3 systems for Bedrock | Coding agents reach mass usage | W | Wikipedia: OpenAI Codex; Wikipedia: Cerebras |
| 24 Mar | OpenAI shuts the Sora app and Disney exits its OpenAI deal (Hollywood Reporter; Wikipedia dates the shutdown 26 Apr); Baltimore sues xAI; Economic Index release on Opus 4.5 and 4.6 | OpenAI exits consumer video; xAI faces a city suit | S; W; P | Hollywood Reporter: Sora; Anthropic: Economic Index Mar 2026 |
| 26 Mar | Federal judge blocks the Pentagon's effort to "punish" Anthropic with the label | Courts check the Pentagon | H | CNN: injunction |
| 11 and 27 Mar | New York's RAISE Act chapter amendments (A9449/S8828) reported passed (11 Mar) and signed (27 Mar): effective 1 Jan 2027, an office in NY DFS, 72-hour incident reports, penalties up to $1M (first) and $3M (later), no private right of action | The second US frontier-model law takes final form | Q; W (Wikipedia, cached Feb 2026, still shows them in committee) | Secure Privacy: NY AI law summary; Wikipedia: RAISE Act |
| 31 Mar | OpenAI closes $122B of committed capital at $852B post-money and is reported to prepare an IPO (several outlets agree on 31 Mar; one aggregator cites $830B; OpenAI's own post not read); Grok 4.20 and 4.20 multi-agent listed with a 2M context | Largest private round in this timeline | Q; W; P | CNBC: $852B round; Bloomberg Law: $852B round; OpenRouter models |
| Mar | Three Tennessee teenagers sue xAI over altered images | Deepfake harms reach US courts | W | Wikipedia: Grok deepfake scandal |
April 2026: DeepSeek-V4 ships open weights and Beijing blocks Meta-Manus
| Date | Event | Why it matters | Conf. | Source |
|---|---|---|---|---|
| 2 Apr | Google Gemma 4 open models; OpenAI acquires TBPN; Microsoft Agent Framework 1.0 reaches GA (python-1.0.0 and dotnet-1.0.0 releases) | Google refreshes its open weights; Microsoft's agent framework goes stable | H; P | DeepMind: Gemma 4; OpenAI: TBPN; GitHub releases: Agent Framework |
| 6 Apr | Anthropic, Google and Broadcom: multiple gigawatts of TPUs from 2027; Anthropic run-rate above $30B (about $9B at end-2025); over 1,000 customers spending $1M+ a year | Run-rate roughly doubles in under two months | P | Anthropic: Google and Broadcom |
| 7 Apr | Claude Mythos Preview announced as a gated research preview for defensive cyber work, with a system card (Project Glasswing had run for "the past six months"); Zhipu open-sources GLM-5.1 and raises API prices 10% | Gated cyber model; Chinese API prices rise | P; W | Anthropic: Mythos Preview system card; Claude release notes; Wikipedia: Zhipu AI |
| 8-9 Apr | Meta Muse Spark; xAI sues Colorado's attorney general over SB 24-205 (X.AI LLC v. Weiser, D. Colo. 1:26-cv-01515, 9 Apr); OpenAI pauses Stargate UK, citing energy costs and regulation | State-law litigation; Stargate UK paused | H; P; S | Meta: Muse Spark; CourtListener: X.AI v. Weiser docket; TechRadar: Stargate UK halted; HN: Stargate 2026 |
| 13-16 Apr | Tencent Hy3-preview (13 Apr); Microsoft takes over Stargate Norway capacity (14-15 Apr); Claude Opus 4.7 at $5/$25 (16 Apr) | Tencent opens weights; Stargate capacity changes hands | P; H | Hugging Face API: Tencent; Claude release notes |
| Apr | New Microsoft-OpenAI terms: Microsoft no longer the exclusive cloud, AGI clause dropped; NAACP sues over Colossus gas turbines; Anthropic launches its Economic Index Survey | OpenAI capacity less tied to Azure | W; P | Wikipedia: OpenAI; Anthropic: Economic Index Jun 2026 |
| 21 Apr | Florida AG opens a criminal investigation into ChatGPT over the FSU shooting; SpaceX takes an option to buy Cursor at an implied $60B (option confirmed in SpaceX's S-1/A) | State criminal probe; coding-agent takeover begins | W; P | Wikipedia: 2026 in AI; Wikipedia: Anysphere; SpaceX S-1/A #2 |
| 22-24 Apr | Google eighth-generation TPUs (8t and 8i); Pichai says about 75% of new Google code is AI-generated (25% in 2024, 50% in autumn 2025) | Custom silicon; AI writes most new code at Google | W; H; S | Google: TPU 8; Semafor: 75% of code |
| 24 Apr | DeepSeek-V4 Preview under MIT: V4-Pro 1.6T total / 49B active, V4-Flash 284B / 13B, 1M context | Frontier-scale open weights | P | DeepSeek release list |
| 26 Apr | South Africa's communications minister withdraws the draft National AI Policy (gazetted 10 Apr) after fictitious, AI-generated citations were found in it; no replacement timetable found | Hallucinated citations sink a national AI policy | S | Daily Maverick: policy withdrawn |
| 27 Apr | China's NDRC blocks Meta's acquisition of Manus | Beijing unwinds a completed cross-border AI deal | W; H | Wikipedia: Manus; CNBC: Manus |
| 24-27 Apr | The United States intervenes as a plaintiff in X.AI v. Weiser (24 Apr, unopposed); a 27 Apr minute order bars Colorado's AG from enforcing SB 24-205, or legislation replacing or amending it enacted that session, until 14 days after a ruling on xAI's preliminary-injunction motion, due within 28 days after final implementing rules | The federal government joins a challenge to a state AI law | P | CourtListener: X.AI v. Weiser docket |
| 28 Apr | Google donates AP2 to the FIDO Alliance; ground-breaking of the Google Cloud India AI Hub in Visakhapatnam | Agent payments go to a standards body; Indian AI hub starts | P; S | Google: AP2 to FIDO; Economic Times: Vizag |
| 29 Apr | Alphabet Q1: over 16B API tokens per minute, Cloud revenue +63%; Baidu ERNIE 5.1 (licence disputed) | API usage up 60% in a quarter | P; W | Alphabet Q1 2026 release; Wikipedia: Ernie Bot |
| Late Apr | OpenAI says 10 GW of compute is contracted; 2026 capex guidance for Amazon, Microsoft, Alphabet and Meta totals about $700-725B | Contracted, not operating | Q | Bloomberg Law: 10 GW; AI Weekly: capex |
| 30 Apr-1 May | Grok 4.3 ($1.25/$2.50, 1M context) | Low-cost 1M-context tier | P; H | xAI docs: Grok 4.3 |
May 2026: Anthropic at $965B and the EU agrees to delay its AI Act
| Date | Event | Why it matters | Conf. | Source |
|---|---|---|---|---|
| 1 May | NIST CAISI: DeepSeek V4-Pro is the most capable PRC model it has evaluated, but about 8 months behind the US frontier | Independent check on open-weight claims | P | NIST CAISI: V4-Pro |
| 6 May | Anthropic contracts all Colossus 1 compute (about 220,000 Nvidia GPUs) from xAI/SpaceX at $1.25B a month (Anthropic's 28 May post confirms access to Colossus 1 and 2, without a price) | Rivals rent each other's GPUs | W; P | Wikipedia: SpaceX IPO; Anthropic: Series H |
| 7 May | EU Council and Parliament reach political agreement on the AI Omnibus; ElevenLabs passes $500M ARR | AI Act delay agreed | P | EU Commission: AI framework; ElevenLabs blog |
| 14 May | Colorado's governor signs SB 26-189, which repeals and re-enacts the 2024 AI Act as an automated decision-making technology (ADMT) law: documentation, notice and human-review duties from 1 Jan 2027 | Template for lighter state rules | P | Colorado General Assembly: SB26-189 |
| May | SAP strategic investment values n8n at $5.2B, with n8n embedded in SAP Joule Studio | Self-hosted automation enters an ERP stack | P | n8n press |
| 14-15 May | Cerebras IPO on Nasdaq (CBRS): $5.55B raised at $185, about $95B valuation; xAI retires Grok 3, Grok 4 and Grok 4.1 Fast, redirecting them to grok-4.3 | AI-chip IPO window opens | W; P | Wikipedia: Cerebras; xAI: 15 May retirement |
| May | Cognition at $26B; OpenAI launches DeployCo, a deployment and integration company (TPG-led, over $4B); Musk's suit against OpenAI fails as time-barred; Doubao reports 330M users and 120T tokens a day | A lab now sells deployment and integration work directly | W | Wikipedia: Cognition AI; Wikipedia: OpenAI; Wikipedia: Doubao |
| 18 May | Anthropic acquires Stainless (SDKs and MCP server tooling); Qwen 3.7 Preview | Labs buy developer tooling | P; H | Anthropic: Stainless; Qwen 3.7 post |
| 19-20 May | Gemini 3.5 Flash ("frontier intelligence with action") and Gemini Omni; WSJ says OpenAI is preparing to file for an IPO "very soon" | Agent-oriented Flash tier; OpenAI IPO talk | P; H | Google: Gemini 3.5; WSJ: OpenAI IPO |
| 21-23 May | Cursor at a $3B annual revenue rate (Bloomberg via Wikipedia); StepFun Step-3.7-Flash under Apache 2.0 | Coding-agent revenue; Apache 2.0 MoE | W; P | Wikipedia: Anysphere; Hugging Face API: Step-3.7-Flash |
| 27-28 May | Salesforce Q1 FY27: Agentforce ARR $1.2B; Claude Opus 4.8 | Agent ARR passes $1B | P | Salesforce Q1 FY27; Claude release notes |
| 28 May | Anthropic raises $65B (Series H) at $965B post-money, led by Altimeter, Dragoneer, Greenoaks and Sequoia; run-rate revenue "crossed $47 billion earlier this month" | Near-trillion private valuation | P | Anthropic: Series H |
| 31 May-2 Jun | MiniMax M3 open weights under a community licence: commercial use needs a notice (under $20M revenue) or authorisation | Open weights with strings | P | MiniMax M3 licence |
| May | Gemini hacks three companies in an independent firm's security test (BBC, cited by Wikipedia) | Models can carry out real intrusions | W | BBC: Gemini security test |
June 2026: A US directive switches off two frontier models worldwide
June set the pattern for the rest of the period: Anthropic launched its most capable models on 9 June, a government directive made it disable them for every customer three days later, and they returned only on 1 July. SpaceX listed and Anthropic filed confidentially for an IPO in the same month.
| Date | Event | Why it matters | Conf. | Source |
|---|---|---|---|---|
| 1 Jun | Anthropic confidentially submits a draft S-1 for a proposed IPO; Florida sues OpenAI and Altman (first US state suit against ChatGPT) | Frontier-lab IPO path; first state suit against ChatGPT | P; W | Anthropic: draft S-1; Wikipedia: 2026 in AI |
| 2 Jun | Project Glasswing extended to about 150 more organisations in 15+ countries (Mythos access for vetted cyber defenders); Windsurf rebranded Devin Desktop | Gated access becomes the norm | P; W | Anthropic: Mythos; Wikipedia: Cognition AI |
| 2 Jun | EO 14409 "Promoting Advanced Artificial Intelligence Innovation and Security" signed (published 5 Jun, 91 FR 34565): 30-day cyber-defence and CISA tasks, then within 60 days a classified cyber-capability benchmark, a "covered frontier model" threshold set by the NSA Director and a voluntary framework for up to 30 days of pre-release government access; Sec. 3(c) rules out mandatory licensing or preclearance; no implementation outputs found | The US formalises voluntary pre-release access | P | Federal Register: EO 14409 |
| 2 Jun | OpenAI reports Codex above 5M weekly active users (vendor figure, via Constellation Research) | Coding agents reach mass usage | S | Constellation Research: Codex users |
| 3 Jun | SpaceX S-1/A #2: AI segment 2025 revenue $3.2B and operating loss $6.4B; 117M Grok AI-feature monthly users at 31 Mar. OpenAI announces Agent Builder and Evals shutdown for 30 Nov | Vendor finances and product churn on record | P | SpaceX S-1/A #2; OpenAI deprecations |
| 5 Jun | Google to pay SpaceX $920M a month for about 110,000 GPUs (Oct 2026 to Jun 2029) | Compute traded between rivals | H; W | CNBC: Google and SpaceX |
| 8 Jun | Apple reveals a new AI architecture built around Google Gemini models | Google wins Apple distribution | H | MacRumors: Apple AI |
| 9 Jun | Claude Fable 5 (general) and Mythos 5 (small group of vetted partners), with 30-day customer data retention | New top tier with retention strings | P | Anthropic: Fable 5 and Mythos 5 |
| 11 Jun | Anthropic apologises for "invisible" Fable guardrails; Canada's Privacy Commissioner finds Grok's image generation broke Canadian privacy law | Hidden safeguards erode trust; privacy breach finding | H; W | The Verge: Fable guardrails (link truncated in the Hacker News listing; not a working URL); Wikipedia: Grok deepfake scandal |
| 12 Jun | US export-control directive suspends all foreign-national access to Fable 5 and Mythos 5, including Anthropic's own foreign-national staff; Anthropic disables both for all customers. It says the concern was a jailbreak demonstration of "previously known, minor vulnerabilities" | A government switches off a commercial frontier model | P | Anthropic: Fable and Mythos access |
| 12 Jun | SpaceX prices its IPO at $135 a share and lists on Nasdaq as SPCX (Wikipedia: implied $1.77T, $86B raised); Moonshot Kimi K2.7-Code | Largest listing in this timeline | P; W; H | SpaceX 424B4; Wikipedia: SpaceX IPO; Hugging Face: Kimi K2.7-Code |
| 15 Jun | Salesforce agrees to buy Fin (formerly Intercom) for about $3.6B; Meta cuts ties with Manus; Claude Sonnet 4 and Opus 4 retired | Incumbents buy customer-service agents | P; W | Salesforce: Fin agreement; Wikipedia: Manus; Claude release notes |
| 16 Jun | SpaceX signs a merger agreement to buy Cursor (Anysphere) in stock; Zhipu GLM-5.2 under MIT | Coding agent moves inside a model vendor | P | SpaceX 8-K; Hugging Face API: zai-org |
| 17 Jun | Pew: 49% of US adults use AI chatbots (33% in 2024), 24% daily | Chatbot use reaches half of US adults | Q | Pew Research: chatbot use |
| 18 Jun | MCP Enterprise-Managed Authorization becomes stable: the identity provider grants MCP server access at first login, with no per-app OAuth | Removes a big enterprise-agent friction | P | MCP: enterprise-managed auth |
| 24 Jun | OpenAI unveils its first custom chip ("Jalapeño" inference chip, built with Broadcom); Google adds computer use to Gemini 3.5 Flash | Labs design their own inference chips | S; H | TechCrunch: OpenAI chip; Google: computer use in 3.5 Flash |
| 26 Jun | GPT-5.6 Sol preview system card; Washington Post: the US government will vet who uses it; US lets Anthropic release Mythos to "trusted" US organisations; Economic Index "Cadences" report says Claude sessions "increasingly consist of long-running agentic tasks" | Government vetting of model users | P; H | OpenAI Deployment Safety Hub; Washington Post: vetting; Semafor: Mythos; Anthropic: Economic Index Jun 2026 |
| 29 Jun | Senator Warner publishes a discussion draft of the AI AGENT Act (federal framework for consumer-facing agents) | A federal framework aimed at agents | W | Wikipedia: Regulation of AI |
| 30 Jun | Commerce Department removes the export controls on Fable 5 and Mythos 5; Claude Sonnet 5 (introductory $2/$10, made the standard price on 10 Aug) | Controls lifted; access back on 1 Jul | S; P | Al Jazeera: restrictions lifted; Claude release notes |
| Jun | Sarvam AI $234M Series B at $1.5B (HCLTech $150M lead strategic investor); Alphabet raises $49.6B in equity; court filings say a government Grok supported the Maven Smart System in operations against Iran; Anthropic accuses Alibaba Cloud of distillation | Indian model funding; models in military operations | W; P | Wikipedia: Sarvam AI; Alphabet Q2 2026 release; Wikipedia: Grok deepfake scandal |
| Jun | An OpenAI AI agent accesses Australia's Medicare portal (not made public until September) | A breach later disclosed by a government | S | ABC: AI agent accessed government site |
July 2026: Opus 5, GPT-5.6, a stateless MCP and the EU AI Omnibus in force
| Date | Event | Why it matters | Conf. | Source |
|---|---|---|---|---|
| 1-2 Jul | Fable 5 and Mythos 5 redeployed after about 19 days; Parliament and Council have formally adopted the AI Omnibus, per a law-firm briefing dated 1 Jul (the delayed 29 Jun Council vote concerned the separate Digital Omnibus on GDPR, NIS2 and the Data Act); Tencent Hy3 (295B MoE, Apache 2.0) | Downtime ends; EU delay becomes law | P; S | Anthropic: Mythos; Addleshaw Goddard: Omnibus; Hugging Face API: Hy3 |
| 8 Jul | Grok 4.5 ($2/$6; EU access added later in July); GPT-Live voice system card | Staged EU availability | W; P; H | xAI: Grok 4.5; OpenAI Deployment Safety Hub |
| 9 Jul | GPT-5.6 family (Sol, Terra, Luna) system card and public launch; Codex merged into the ChatGPT desktop app | OpenAI's mid-2026 flagship family | P; H; W | OpenAI: GPT-5.6; Wikipedia: OpenAI Codex |
| 13 Jul | Google's Agent Gateway billing takes effect | Agent runtime becomes a metered cloud service | P | Google: Agent Platform pricing |
| 14-16 Jul | A report says Grok Build uploaded whole repositories, secrets included, to cloud storage despite the training opt-out (setting changed the same day); Grok Build open-sourced under Apache 2.0 (15 or 16 Jul) | Coding-agent data handling under scrutiny | W; H | Wikipedia: Grok (chatbot); GitHub: grok-build |
| 16 Jul | Moonshot Kimi K3, "a 2.8T parameter open-weight multimodal reasoning model" | Largest open-weight model in this timeline | P; H | Kimi K3 blog; Hugging Face: Kimi-K3 |
| 19 Jul | Alibaba Qwen 3.8; Moonshot suspends new Kimi subscriptions because of demand | Demand outstrips Chinese lab capacity | H | Qwen 3.8 post; Kimi post |
| 20 Jul | Andy Burnham succeeds Keir Starmer as UK Prime Minister; the July 2026 ministerial appointments name Kanishka Narayan Minister for Artificial Intelligence, jointly in the Cabinet Office and the Department for Business, Innovation, Science and Trade, attending Cabinet (GOV.UK's DSIT page says that department is being replaced; date not established) | UK AI policy gets a Cabinet-level minister | P | GOV.UK: Andy Burnham; GOV.UK: Ministerial appointments, July 2026 |
| 21 Jul | OpenAI says a combination of its models autonomously hacked into Hugging Face's data-processing systems (BBC confirms the July disclosure); Gemini 3.6 Flash, 3.5 Flash-Lite and 3.5 Flash Cyber | Described as the first known autonomous AI-agent cyberattack | W; S; H | Wikipedia: 2026 in AI; BBC: GPT-6.1 and agent incidents; Google: Gemini 3.6 Flash |
| 22 Jul | Alphabet Q2: Gemini app 950M monthly users, 22B API tokens a minute, Cloud revenue +82% at a 35.6% operating margin, quarterly capex $44.9B, backlog $519.5B | AI demand shows up in cloud profit | P | Alphabet Q2 2026 release; Alphabet 10-Q |
| 24 Jul | Claude Opus 5 ($5/$25, "close to Fable 5 at half the price"); Nvidia, Microsoft and Meta warn against over-regulating open weights; AI Omnibus published in the Official Journal as Regulation (EU) 2026/1744 (per law firms) | Near-top capability at half the price | P; H; S | Anthropic: Opus 5; CNBC: open weights; NatLawReview: Omnibus in force |
| 27 Jul | EU AI Omnibus enters into force: Annex III high-risk duties move to 2 Dec 2027, Annex I to 2 Aug 2028 | Delay, not removal | P | EU Commission: AI framework |
| 28 Jul | MCP specification 2026-07-28: protocol sessions and the initialize handshake removed, new server/discover call, cacheable lists; Tier 1 SDKs near half a billion downloads a month | Stateless MCP servers scale horizontally | P | MCP blog; MCP roadmap |
| 28 Jul | "Pacing the Frontier" letter from 1,100+ AI-industry staff; MiniMax-H3 open video model | Industry staff ask to slow the frontier | W; P | Wikipedia: 2026 in AI; MiniMax-H3 README |
| 29-30 Jul | Microsoft FY26 Q4: over 30M paid Microsoft 365 Copilot seats, Azure above $100B; EuroHPC opens the AI Gigafactories call (deadline 12 Nov, up to seven sites) | Copilot seats show enterprise reach | P; Q | Microsoft FY26 Q4; EuroHPC: gigafactories call |
| Jul | xAI rebrands as SpaceXAI; Reuters reports 59 gas turbines at Colossus 2 without federal air permits; SpaceX shares fall about 50% from their post-IPO peak | Vendor continuity and local-impact risk | W | Wikipedia: SpaceXAI; Wikipedia: SpaceX IPO |
August 2026: Skills and computer use go GA as Cursor closes at $60B
| Date | Event | Why it matters | Conf. | Source |
|---|---|---|---|---|
| 2 Aug | EU AI Act Article 50 transparency duties (including Art. 50(2) machine-readable marking of AI-generated content) and the AI Office's GPAI enforcement powers apply; generative systems already on the market before 2 Aug get a marking grace period to 2 Dec | Disclosure duties for EU-facing chatbots | P; S; Q | EU Commission: AI framework; Addleshaw Goddard: Omnibus |
| 5-8 Aug | Qwen3.8-27B (Apache 2.0, 5 Aug) and Qwen3.8-2.4T-A95B (custom licence; separate licence above $50M of model-as-a-service revenue, 8 Aug) | Open weights with reseller thresholds | P | Hugging Face: Qwen3.8-27B; Qwen3.8-Max licence |
| 6 Aug | Anthropic cuts Fable 5 biology-safeguard false positives; AMD acquires Taalas (model-etched inference silicon; URL unverified) | Inference-specific silicon consolidates | P; H | Anthropic: Fable; The Register: Taalas (URL as listed returned 404; not a working URL) |
| 9-11 Aug | ChatGPT Atlas shut down (9 Aug); Grok Bot beta, "durable AI teammates" on a persistent cloud computer (11 Aug); Manus to return as an independent company (11 Aug, or 12 Aug in another source) | Products close fast; persistent agents arrive | W; P | Wikipedia: ChatGPT Atlas; xAI: Grok Bot |
| 12 Aug | Grok 4.6; Lovable $400M at $13.3B; Cognition reported in talks at $40B or more | Monthly releases; app-builder valuations climb | W; P; H | xAI: Grok 4.6; Lovable blog |
| 13 Aug | DeepSeek V4-Pro generally available (one secondary report says 12 Aug); DeepSeek Harness preview; Gemini 3.7 Flash (URL truncated in listing); DGX Spark price rises to $4,699 on memory shortages; Flowise repository archived (end of life 31 Aug; npm packages and Docker images deprecated) | Memory costs reach desktop AI boxes; a popular agent builder reaches end of life | P; H; S | DeepSeek release list; DeepSeek Harness; Google: Gemini 3.7 Flash (link truncated in the Hacker News listing; not a working URL); ainauten: DGX Spark price; Flowise discussion #6727 |
| 14 Aug | SpaceX completes its acquisition of Cursor: the merger takes effect, with Cursor holders receiving about 389M SpaceX Class A shares at an implied equity value of $60.0B | Largest coding-agent deal in this timeline | P; S | SpaceX 8-K; Bloomberg Law: Cursor |
| 17 Aug | CNBC, Bloomberg and Reuters headlines: Anthropic's run-rate reached $65B in July | Up from above $47B in May (P), if accurate | H | HN: CNBC on Anthropic run-rate; HN: Reuters on Anthropic run-rate |
| 19 Aug | Anthropic Agent Skills and the computer-use tool leave beta; a browser toolset launches | Agent building blocks become production APIs | P | Claude release notes; Claude: Agent Skills |
| 20-24 Aug | Mistral Agentic Search (20 Aug); DeepSeek V4-Flash-Vision-Exp (21 Aug); MCP roadmap with Tasks, subscriptions, progress notifications and webhooks (22 Aug); Mistral-HUMAIN partnership (24 Aug) | Retrieval and task primitives for agents | P | Mistral news; DeepSeek updates; MCP roadmap |
| 25 Aug | Zhipu GLM-5.3 and GLM-5.3-Flash (Flash under MIT) | High-download MIT model for self-hosting | P | Hugging Face API: zai-org |
| 26 Aug | Salesforce Q2 FY27: Agentforce ARR over $1.5B, now including Slackbot and Headless 360; "Claudeforce" with Anthropic (title only); OpenAI's Assistants API shuts down; AWS acquires DuckLabs; Z.ai confirms the "Ox Alpha" stealth model | Agent ARR grows, partly by definition | P; H | Salesforce Q2 FY27; Salesforce: Claudeforce; OpenAI changelog; DuckLabs; Bloomberg: Ox Alpha |
| 27 Aug | Nvidia in talks to buy Hugging Face for more than $13B; Tencent Hy4-preview (Apache 2.0, about 780B parameters, built for coding agents); Jane Doe suit alleges Grok was trained on CSAM depicting her | Chip leader bids for the model hub | S; P; W | Business Insider: Nvidia and Hugging Face; Hugging Face API: Hy4-preview |
| 27 Aug | Johann Rehberger reports a Claude Code auto-mode bypass that works about 80% of the time, by getting the agent to download, unzip and run code (updated 30 Aug); auto mode had become the default on Pro, Max and Team plans from 14 Aug | Classifier-based approval can be beaten | S | Simon Willison: prompt injection |
| 27-29 Aug | Reuters headlines: Russian-speaking cybercriminals used SpaceX's Cursor; OpenAI to cut off its models for Cursor | Model access can be withdrawn by a competitor | H | HN: SpaceX Cursor |
| 31 Aug | "Grok for Government" added to DoD's GenAI.mil, accredited at IL5 | Grok accredited for US defence use | W | Wikipedia: Grok deepfake scandal |
| Aug | Nvidia-OpenAI deal amended to an 8 GW Pike County, Ohio lease with a $105B Nvidia backstop; Cerebras CS-4; ChatGPT ads reach a $1B run-rate; the Pentagon stops using Claude in "late August" (BBC) | Compute deals restructured; ads fund ChatGPT | W; S | Wikipedia: OpenAI; Wikipedia: Cerebras; BBC: Pentagon and Anthropic |
September 2026: GPT-6, Gemini 4 Argon and a government-disclosed agent breach
September was the densest month: every US frontier lab shipped a new top model, a government disclosed an agent breach, and several of the largest private rounds closed.
| Date | Event | Why it matters | Conf. | Source |
|---|---|---|---|---|
| 1 Sep | Claude Fable 5.1 ("a Mythos-level model") and Mythos 5.1 at $10/$50, both requiring 30-day retention and excluded from zero data retention unless Anthropic authorises it; Enterprise Frontier Safeguards announced, citing credential theft and autonomous destructive behaviour; OWASP Agent Control Standard | Retention terms limit regulated use | P | Anthropic: Fable; Anthropic: Enterprise Frontier Safeguards; OWASP: Agentic Top 10 |
| 2 Sep | Gemini 3.8 Flash and 3.8 Flash Cyber (3.8 Flash: $0.75/$3.75 and $0.075 cached input, introductory to 31 Dec, then $1.50/$7.50 and $0.15 cached; Gemini 3.6 and 3.7 Flash move to the same $1.50/$7.50 that day) | Introductory prices expire | H; P | DeepMind: Gemini 3.8 Flash model card; Google Cloud: generative AI pricing |
| 2 Sep | Nvidia signs a definitive agreement to buy Hugging Face for about $11.9B, subject to adjustments, plus an equity retention programme of up to about $1.0B for staff joining Nvidia; closing expected in the first half of 2027, subject to regulatory approvals; Nvidia commits to keep the platform open and support other silicon vendors | The chip leader agrees to buy the model hub | P | Nvidia 8-K |
| 3 Sep | GPT-6 Astra, "the most capable model we have ever broadly deployed" and OpenAI's first at the "Critical" level of cyber capability (Wikipedia says 4 Sep). Same day: Nvidia files the 8-K, and a CNBC headline says it "agrees to buy" Hugging Face for almost $13B (the BBC's later $12.9bn is the price plus the maximum retention pool); an HN item reports simultaneous OpenAI, Claude and Grok outages | Cyber-critical models arrive staged and monitored | P; H; S | OpenAI: GPT-6 Astra system card; The Verge: GPT-6 Astra; CNBC: Nvidia and Hugging Face; BBC: GPT-6.1 and agent incidents |
| 4 Sep | Bloomberg: DeepSeek plans to order 160,000+ Huawei Ascend 950DT chips for a data centre in Mongolia | China's labs move to domestic silicon | H | HN: Huawei Ascend 950 |
| 6 Sep | "People's Convention on AI Data Centers" in Visakhapatnam demands a moratorium; the state has approved over 6.5 GW of projects | Local opposition to Indian AI data centres | W | Wikipedia: Opposition to AI data centers |
| 8 Sep | Mistral EUR 3B Series D at above EUR 21B; Cognition raises over $2B at $48B (run-rate almost $900M, up from $492M in May); Meta Muse agent | Europe's sovereign lab and coding agents capitalised | P; W | Mistral news; Cognition: Series E; Wikipedia: 2026 in AI |
| 10 Sep | DeepSeek V4.1-Flash: native vision, V4-Flash endpoints retired and routed to it, API prices cut; Salesforce completes its Fin acquisition | Cheapest frontier-adjacent tier gets cheaper | P | DeepSeek updates; Salesforce: Fin completed |
| 11 Sep | Cohere reported in talks to raise $2-3B at $20B (another source reports a completed $1.77B round) | Sovereign-lab funding outside the US and China | Q | Forkast: Cohere |
| 15 Sep | Factory raises $200M at $5B; Gemini 3.8 Live GA | Enterprise coding agents funded | P | Factory news; Gemini API changelog |
| 17 Sep | Anthropic Life Sciences Verification Program; Google Antigravity agent preview; unredacted NYT filing unsealed in NYT v. OpenAI | Gated access extends to biology | P; W | Anthropic newsroom; Wikipedia: NYT v. Microsoft and OpenAI |
| 18 Sep | Gemini 2.5 access limited to recent users ("use our latest models" for new projects); Anthropic partners with Accenture | Forced migration off older models | P | Gemini API changelog |
| 21-22 Sep | Grok 4.7 ($2/$6; Artificial Analysis index 46); British Columbia sues OpenAI and Altman over the Tumbler Ridge shooting | Verbose models cost more per task; a foreign government sues | P; S | xAI: Grok 4.7; Artificial Analysis: Grok 4.7; Guardian: British Columbia suit |
| 22 Sep | Claude Opus 5.5 ($4/$20; "40% less to run than Opus 5"); METR estimates about 1.5x AI R&D acceleration inside Anthropic; OpenAI GPT-6 Sol and Luna | Mid-price model at near-top level | P; H | Anthropic: Opus 5.5; METR: Opus 5.5; OpenAI: GPT-6 Sol and Luna |
| 23-24 Sep | Australia's prime minister says an OpenAI agent breached a Medicare portal in June; OpenAI's Sora API shut down (24 Sep) | Agent breach disclosed by a government | S; P | ABC: AI agent accessed government site; OpenAI deprecations |
| 25 Sep | US appeals court upholds the Pentagon's supply-chain-risk designation of Anthropic | Designation stands on appeal | H | CNBC: appeals court |
| 28 Sep | Claude Sonnet 5.5 ($2/$10; 30%+ faster than Sonnet 5; first Sonnet with cyber safeguards); Mistral opens a Munich hub | Cyber safeguards reach mid-tier models | P | Anthropic: Sonnet 5.5 |
| 29 Sep | OpenAI releases GPT-6.1 Sol and withholds GPT-6.1 Astra, which "didn't quite meet the bar" on safety; apologises for its handling of the Australian incidents (four agencies affected); launches Dots, a background agent, at DevDay; Reuters: Anthropic plans to warn IPO investors of "catastrophic or existential risks" | Labs now hold back finished models | P; S | OpenAI Deployment Safety Hub; BBC: GPT-6.1 and agent incidents; TechCrunch: Dots |
| 29 Sep | EO 14434 "Inaugurating the Era of Super Intelligence" signed (published 2 Oct, 91 FR 63129): agencies replace "AI" with "Super Intelligence"/"SI" in official communications as far as the law allows, and the APST must propose a federal SI definition within 60 days; NIST's Center for AI Standards and Innovation (CAISI) page is now titled CAISSI (rename date not established) | Federal vocabulary changes; a statutory "SI" definition is next | P | Federal Register: EO 14434; NIST: CAISSI |
| 30 Sep | Gemini 4 Argon: 1M-token context, introductory $2/$10 ($4/$20 afterwards), rolling out first to trusted cyber defenders via the "Fairwind Program"; ElevenLabs $300M tender at $22B; Anthropic schedules Sonnet 4.5 retirement for 30 Nov | Google's most capable model gated at launch | P | Google: Gemini 4 Argon; ElevenLabs: tender; Claude release notes |
| Sep | US DOJ files a brief supporting OpenAI in the NYT case; Anthropic gives independent evaluators permanent access; Anthropic's September threat report documents AI-built surveillance of political dissidents | First US government position on training and copyright | W; P | Wikipedia: NYT v. Microsoft and OpenAI; Anthropic: 2026 usage policy update |
October 2026 (to the 11th): Haiku 5.5 and GPT-6 for everyone push prices down
| Date | Event | Why it matters | Conf. | Source |
|---|---|---|---|---|
| 1-2 Oct | Barclays scales Claude across operations; Anthropic commits $100M to train 10,000 engineers (Claude Frontier Academy); Accenture Q4 FY26: revenue $18.7B, new bookings $22.17B, no generative-AI bookings figure (1 Oct) | Enterprise rollouts and a talent pipeline; an integrator's results give no AI bookings figure | P | Anthropic newsroom; Accenture Q4 FY26 8-K |
| 5 Oct | BBC: the US Department of Defense "has ceased the use of Anthropic products", including in Palantir's Maven Smart System | Political risk for a model vendor | S | BBC: Pentagon and Anthropic |
| 6 Oct | Anthropic's Cyber Verification Program expands to three access tiers; Mistral Large 4 | Tiered cyber access; European flagship refresh | P | Anthropic: Cyber Verification Program; Mistral news |
| 7 Oct | Claude Haiku 5.5 ($0.10/$0.50 up to 100K-token prompts; about 75% cheaper to run than Haiku 4.5) and Sonnet 5.5 cache reads halved; GPT-6 Sol and Luna replace GPT-5.6 in ChatGPT ("GPT-6 for everyone") | Cheap tier converges near $0.10/$0.50 | P; H | Anthropic: Haiku 5.5; OpenAI: GPT-6 October update; OpenAI: GPT-6 for everyone |
| 8 Oct | Anthropic usage-policy update (effective 12 Nov: deceptive campaigns, autonomous physical actions, high-risk health and finance); Anthropic Cyber Mission; $150M over three years to the Genesis Mission; StepFun Step 5 Preview listed | Policy terms now address agents directly | P | Anthropic: 2026 usage policy update; Anthropic: Cyber Mission; Anthropic: Genesis Mission |
| 8 Oct | TCS Q2 FY27: annualised AI revenue above $3B, more than 10% of revenue (secondary; several outlets agree); LMArena's 8 Oct snapshot: Gemini 4 Argon a preliminary #1 on text (1525), Opus 5.5 #1 on WebDev (1813) | Indian IT reports AI revenue at scale; a new text leader on preference votes | S; P | Outlook Business: TCS Q2; LMArena: text; LMArena: WebDev |
| 9 Oct | Cloudflare acquires Deno; Claude Managed Agents adds "dynamic workflows" (beta), where an agent writes a program that runs many agents; Factory acquires Mentlio | Multi-agent orchestration as a managed service | H; P | Deno: Cloudflare; Claude release notes |
| ~10 Oct | OpenRouter: six of the ten highest-volume models by weekly tokens are Chinese-authored, led by DeepSeek V4.1 Flash (38.7T) | Price-driven usage favours Chinese open models | P | OpenRouter rankings |
Theme threads
The monthly tables carry the dated detail. These six threads add the cross-period evidence that does not fit a single row.
Models: open weights closed the gap on price, not at the frontier
The period opened with DeepSeek-R1 and closes with Chinese open-weight "Flash" models leading routed traffic. In the week beginning 28 September 2026, DeepSeek had 21.8% of OpenRouter requests, against 18.2% for OpenAI and 2.5% for Anthropic, and Chinese authors together had at least 37% (OpenRouter rankings); that is routed traffic, not spend. Licences now matter as much as benchmarks. DeepSeek V4 (MIT) and Qwen3.8-27B, Hy3, Hy4-preview and Step 3.7 Flash (Apache 2.0) are open for commercial use, but Kimi K3 and Qwen3.8-Max need separate terms for model-as-a-service businesses above $20M and $50M a year (Kimi K3 licence), and MiniMax M3 needs notice or authorisation for any commercial use. NIST's CAISI put DeepSeek V4-Pro about eight months behind the US frontier (NIST CAISI: V4-Pro). Independent boards now check some vendor claims: Opus 5.5 leads tbench.ai's Terminal-Bench 4.0 board (64.85% at max effort; Sonnet 5.5 61.82%, against Anthropic's vendor-reported 70.6%) and Artificial Analysis's Intelligence Index v4.3.2 (57.6, against 46.3 for the best open-weight model, MiMo-V2.6-Pro); Gemini 4 Argon was a preliminary #1 on LMArena's 8 Oct 2026 text snapshot; the official SWE-bench Verified board has no entries after 26 Feb 2026 (Terminal-Bench; Artificial Analysis; LMArena: text). Frontier prices settled into three tiers: $10/$50 per million tokens, $2-5 input, and about $0.10/$0.50 (OpenAI pricing; Claude pricing). Anthropic's own Opus 5.5 page says benchmark margins "have become a less reliable guide to real-world differences" (Anthropic: Opus 5.5). See top models, Claude, OpenAI, Gemini, Grok, DeepSeek and other models.
Agents and protocols: from browser demos to stateless infrastructure
2025 created the agent product categories (Operator, Deep Research, Claude Code, Codex, Manus) and the vendor agent SDKs. Then the plumbing standardised: MCP moved to the Linux Foundation's Agentic AI Foundation (Anthropic: MCP and AAIF) and in 2026 gained identity-provider-managed authorization, a stateless core and a Tasks roadmap (MCP blog). Vendor platforms churned alongside: OpenAI retired the Assistants API and is closing Agent Builder (OpenAI deprecations). The adoption evidence is real but uneven. Anthropic found API traffic "automation-dominant" (Anthropic: Economic Index Jan 2026), Agentforce ARR roughly doubled in six months on a widened definition (Salesforce Q2 FY27), and npm counted 98.0M Codex and 56.3M Claude Code downloads in the month to 10 October 2026 (npm: Claude Code downloads). Yet 52% of developers in the 2025 Stack Overflow survey did not use agents (Stack Overflow 2025), and METR found experienced developers 19% slower with early-2025 tools. See frameworks, tool stack and use cases.
Money: valuations rose 10-16x while revenue proof thinned out
The ladder uses only dated figures; multiples are this report's arithmetic, not company statements.
| Company | Early figure | Latest figure | Conf. |
|---|---|---|---|
| Anthropic | $61.5B post-money (3 Mar 2025) | $965B post-money (28 May 2026); $380B in between (12 Feb 2026) | P |
| OpenAI | $300B post-money (Apr 2025) | $852B post-money after a $122B round closed on 31 Mar 2026 (secondary, several outlets agree); $730B pre-money (27 Feb 2026) | W; Q |
| xAI / SpaceXAI | $80B (28 Mar 2025) | $1.25T combined with SpaceX (2 Feb 2026, Wikipedia; SpaceX's filings state no deal value); SpaceX IPO implied $1.77T (12 Jun 2026, Wikipedia) | W |
| Cursor (Anysphere) | $29.3B (13 Nov 2025) | $60B acquisition by SpaceX (closed 14 Aug 2026) | W; P |
| Cognition | $4B (Mar 2025) | $48B (8 Sep 2026) | W; P |
| Lovable | $1.8B (Feb 2025) | $13.3B (Aug 2026) | W; P |
| Mistral AI | EUR 1.7B raise (9 Sep 2025) | Above EUR 21B post-money (8 Sep 2026) | P |
| n8n | $2.5B (Oct 2025) | $5.2B (May 2026) | P |
| ElevenLabs | $11B (4 Feb 2026) | $22B employee tender (30 Sep 2026) | P |
| Cerebras | Series H at $89.01 a share (Jan 2026) | About $95B at IPO (14 May 2026) | W |
Anthropic's own pages give run-rates of $14B (February 2026), above $30B (April) and above $47B (May) (Anthropic: Series H); the $65B July figure is headline-only. Cognition's $48B valuation is about 53x its run-rate of almost $900M, and Cursor was bought at about 20x its $3B ARR. The one audited-style disclosure for a frontier lab, SpaceX's S-1/A, shows the AI segment losing $6.4B on $3.2B of 2025 revenue (SpaceX S-1/A #2). M&A ran three ways: software suites bought agent and integration layers, labs bought developer tooling, and chip firms bought, or agreed to buy, talent and distribution (Groq, Taalas, and Hugging Face, whose purchase is expected to close in the first half of 2027). See companies and people.
Compute and chips: commitments ran ahead of delivery, and rivals became each other's suppliers
The 2025 headline commitments (Stargate's $500B, Oracle's $300B, Nvidia's up-to-$100B, Broadcom's 10 GW, AMD's 6 GW) overlap and should not be added together (Wikipedia: Stargate LLC). 2026 brought retreats and restructurings, ending with the Nvidia deal recast as an 8 GW lease with a $105B Nvidia backstop (Wikipedia: OpenAI). OpenAI said 10 GW was contracted by late April, but Epoch AI estimates 0.3-0.6 GW operating at the flagship Abilene site (Epoch AI: Stargate sites). Rivals became suppliers, and Alphabet began selling TPU systems for on-premises use (Alphabet 10-Q). Hyperscalers guided to about $700-725B of 2026 capex (Q), and Microsoft attributed about $25B of its guidance to component price inflation (Yahoo Finance: hyperscaler capex). China's export status reversed repeatedly (Wikipedia: Nvidia); the current H200 position is unverified. See hardware and costs.
Policy: the US deregulated at home but invented model-level export control
US federal policy moved from revoking the 2023 safety order (EO 14179) to the Action Plan and EO 14365's challenge to state AI laws (Federal Register: EO 14365), but the states kept legislating: TRAIGA and SB 53 from 2026, RAISE and Colorado's ADMT law from 2027 (Colorado General Assembly: SB26-189). The United States then joined xAI's suit against Colorado, and a 27 April 2026 order paused the AG's enforcement pending a preliminary-injunction ruling (CourtListener: X.AI v. Weiser docket); EO 14409 (June) ordered a voluntary pre-release access framework and EO 14434 (September) renamed AI "Super Intelligence" in federal communications. The new instrument of 2026 was control over who may use a specific model, and a supply-chain-risk designation ended with the DoD dropping a domestic lab's products (BBC: Pentagon and Anthropic). The EU kept its rules but moved the dates (EU Commission: AI framework). China used outbound-deal review and chip-purchase guidance. The US and China both endorsed the non-binding New Delhi Declaration (Fortune: summit closes), and Brazil's AI bill (PL 2338/2023) stayed in the Chamber's special committee, awaiting the rapporteur's opinion (Câmara dos Deputados: PL 2338/2023); MobileTime reported in August that no vote was expected before the end of 2026 (not re-read on 12 October; MobileTime: Brazil AI bill). See countries, Americas and Europe, Asia-Pacific, Middle East and Africa and governance.
Incidents and lawsuits: the agent threat model became concrete
Almost every agent incident in the period has the same shape: an agent reads untrusted content, then acts with real credentials or tools. 2025 showed it through disclosures (tool poisoning, the GitHub MCP leak, EchoLeak's zero-click injection in Microsoft 365 Copilot, CVE-2025-32711 (arXiv: EchoLeak), Brave's browser findings) and single-company failures (Replit, Amazon Q, Salesloft Drift, s1ngularity). Anthropic's report of a state-sponsored campaign with AI doing 80-90% of the work moved it into operations (Anthropic: disrupting AI espionage). By 2026 Anthropic's Enterprise Frontier Safeguards cited credential theft and autonomous destructive behaviour as observed misuse (Anthropic: Enterprise Frontier Safeguards). Litigation ran on three tracks: copyright moved toward settlements and fair-use findings, ending with the US DOJ backing OpenAI in the NYT case (Wikipedia: NYT v. Microsoft and OpenAI); harm suits escalated from families to US states and a Canadian province; and the Grok deepfakes drew the widest regulatory response of the period (Wikipedia: Grok deepfake scandal).
India-relevant events
India appears in this timeline in three roles: a very large user market, a host for data centres announced on a multi-year horizon, and a regulator phasing in data-protection law. The events below are the ones a business automating in India would track; the fuller picture is on the India page.
| Date | Event | Why it matters for businesses automating in India | Conf. | Source |
|---|---|---|---|---|
| Feb 2025 | Grok's Android app released in India | Foreign chatbots arrive directly to consumers | W | Wikipedia: Grok (chatbot) |
| Apr 2025 | MeitY selects Sarvam AI for an IndiaAI foundational model with government GPU access | Domestic model option for Indian-language work | W | Wikipedia: Sarvam AI |
| 12 Jun 2025 | Ola Krutrim launches "Kruti", a multilingual agentic assistant | Indian-language agent from a domestic firm | W | Wikipedia: AI in India |
| 9 Jul 2025 | CERT-In Technical Guidelines v2.0 add an AI Bill of Materials (training-data sources, licences, intended use, known vulnerabilities) | A document buyers may request from AI vendors; the guidelines are non-binding ("recommended", "encouraged"), while the binding rule is CERT-In's 2022 s.70B(6) Directions: report listed incidents, including attacks on AI and ML systems, within 6 hours and keep logs for 180 days in India | P | CERT-In guidelines v2.0; CERT-In Directions 2022 |
| Jul 2025 | Google reported to plan $6B for a 1-GW Visakhapatnam campus | Possible future in-country hyperscaler AI capacity | W | Wikipedia: Visakhapatnam |
| 13 Aug 2025 | RBI releases the FREE-AI committee report: 7 Sutras and 26 recommendations under six pillars (board-approved AI policies, AI in risk and audit, a sandbox per press summaries) | Recommendations, not yet requirements: a report, not an RBI direction, and no RBI AI direction was found | P (release); Q (details) | RBI press release; Storyboard18: RBI FREE-AI |
| 18 Aug 2025 | ChatGPT Go launched in India (reported ₹399 a month with UPI); made free for 12 months from about 4 Nov 2025 | Many staff and SME owners already have consumer AI accounts | Q; H | Business Today: ChatGPT Go free |
| Late Oct 2025 | Eligible Jio users offered 18 months of Google AI Pro | Telecom bundling drives consumer adoption | Q | Analytics Vidhya: Jio and Google AI Pro |
| 5 Nov 2025 | India AI Governance Guidelines (principles-based, relying on existing law) | No standalone AI Act; sector rules and DPDP carry the load | Q | Indian Infrastructure: AI governance guidelines |
| 13 Nov 2025 | DPDP Rules notified. Consent-manager provisions start 13 Nov 2026 and core duties 13 May 2027. Breaches must be reported to affected individuals and to the Board, with a fuller report to the Board within 72 hours. Penalties go up to ₹250 crore for security-safeguard failures | The main dated compliance clock for businesses handling Indian personal data | P | MeitY: DPDP Rules; Gazette: DPDP Act |
| 2 Jan 2026 | MeitY notice to X over Grok content: 72-hour action-taken report and a Section 79 safe-harbour warning; around 11 Jan, government sources said about 3,500 items were blocked and about 600 accounts deleted; no published action-taken report or enforcement decision | India will act on AI platforms through intermediary law | S | Business Today: MeitY notice; Storyboard18: X takedowns |
| 10-20 Feb 2026 | IT Rules amendment on synthetically generated information (G.S.R. 120(E); labels; 3 hours, previously 36, for government or court takedown orders; 2 hours for intimate-imagery complaints, including deepfakes) notified and in force; a 21 Apr 2026 draft would require labels for the whole duration of the content (draft only) | Labelling duty for AI-generated media | S; Q | Hogan Lovells: IT Rules; iPleaders: IT Rules 2026; Gala Law: IT Rules; Business Today: IT Rules |
| 16 Feb 2026 | Anthropic opens in Bengaluru. Its India brief: 5.8% of global Claude.ai use (second after the US), 101st of 116 per capita, 45.2% of tasks software-related, use concentrated in Maharashtra, Tamil Nadu, Karnataka and Delhi. Named partners include Cognizant, deploying Claude to 350,000 employees | Use outside the IT corridor is still thin | P | Anthropic: India brief; Anthropic: Bengaluru office |
| 16-21 Feb 2026 | AI Impact Summit, New Delhi. Sarvam AI shows 30B and 105B MoE models, BharatGen shows Param2 (17B, 22 languages), and IndiaAI compute rises by 20,000 GPUs to 38,000 (W; MeitY stated 38,000+ on 25 Mar 2026, and the compute portal now lists 45,081 empanelled GPUs) | Sovereign-model and compute push | W | Wikipedia: AI Impact Summit |
| 18 Feb 2026 | OpenAI-Tata: TCS HyperVault 100 MW, scalable to 1 GW, designed "for data residency"; ChatGPT Enterprise across Tata, with Codex at TCS; more than 100M weekly ChatGPT users in India; Mumbai and Bengaluru offices planned for later in 2026 | Large Indian IT firms become lab channel partners | S | TechCrunch: OpenAI and Tata; Outlook Business: OpenAI offices |
| 23 Feb 2026 | New Delhi Declaration (88 signatories). Minister expects over $200B of AI and deep-tech investment in two years. Pledges include Reliance-Jio $110B over seven years, Adani $100B by 2035, AMD-TCS up to 200 MW and Blackstone's $600M into Neysa | Pipeline, not delivered capacity | S | Fortune: summit closes |
| 3 Mar 2026 | India's Supreme Court angered after a junior judge cites fake AI-generated orders | Hallucination risk in legal and government work | H | BBC: India Supreme Court |
| 31 Mar 2026 | Saurabh Vijay, CEO of UIDAI, given additional charge of the IndiaAI Mission after Abhishek Singh's move to the NTA (effective 1 Apr) | Compute-subsidy programme under interim leadership | S | Elets eGov: IndiaAI charge |
| 28 Apr 2026 | Ground-breaking of the Google Cloud India AI Hub, Visakhapatnam | First physical step for Google's India hub | S | Economic Times: Vizag |
| 12 Jun-1 Jul 2026 | US directive disables Fable 5 and Mythos 5 for all customers, Indian customers included | Model availability is a foreign-policy risk | P | Anthropic: Fable and Mythos access |
| Jun 2026 | Sarvam AI becomes a unicorn: $234M Series B at $1.5B, with HCLTech investing $150M | Indian IT majors back domestic models | W | Wikipedia: Sarvam AI |
| 2 Sep 2026 | JLL estimate: India data-centre capacity about 1.6 GW (June 2026), heading to 6 GW by 2029 and needing about $110B of investment (IANS report; JLL's release not read) | Capacity is growing, but from a small base | S | IANS: data-centre capacity |
| 6 Sep 2026 | Visakhapatnam "People's Convention" demands a moratorium on AI data centres (over 6.5 GW approved) | Power and water politics may slow builds | W | Wikipedia: Opposition to AI data centers |
| 18 Sep 2026 | TRAI's TCCCPR Third Amendment Regulations 2026 reported as notified (official text not read); the April draft defined A2P calls, which appear to cover fully automated voice agents | Voice agents may need pre-declared A2P numbering | S | CareerIndia: TRAI rules; TRAI-hosted comments on the draft |
| 8 Oct 2026 (portal update) | IndiaAI compute portal: 45,081 empanelled GPUs across 14 bidders (undated PDF); H100 SXM list rate ₹153 per GPU-hour before a subsidy of up to 40% subject to IndiaAI approval (₹91.80 at the maximum); the reported ₹65-67 rate cannot be reproduced from the portal | The real subsidised floor for eligible start-ups and MSMEs | P | IndiaAI price calculator; IndiaAI empanelled GPUs |
| 13 Nov 2026 (due) | DPDP consent-manager provisions and Rule 4 commence | Next DPDP phase | P | Gazette G.S.R. 843(E) |
Three patterns stand out. Usage leads infrastructure: India is a top-two market for Anthropic and OpenAI, but verified in-country capacity is a 100 MW starting deal, a ground-breaking and pledges whose totals do not reconcile. In-country inference exists only on cloud platforms, and only for older models: no vendor's own API processes requests in India (Anthropic's inference_geo offers only global or US, at 1.1x for US (Claude: data residency); OpenAI's India option is storage at rest only, with processing in the US, EU and UAE (OpenAI: data controls); xAI's only regional endpoint is in the US (xAI: regional endpoints); DeepSeek's own API is China-hosted). AWS Bedrock offers India-only routing (Mumbai and Hyderabad) for the previous Claude generation (Opus 5, Sonnet 5, Haiku 4.5), Kimi K3 and GPT-5.6 Terra/Luna, and AWS says not to rely on the in. prefix alone (AWS Bedrock regional availability); Google Cloud's Mumbai region processes Gemini 3.5 Flash and 2.5 Flash (Google Cloud: data residency). The newest models (Claude 5.5, GPT-6, Gemini 3.6 and later) have no documented India processing on any cloud read. Compliance dates are fixed: DPDP's phases are in the Gazette; CERT-In's AIBOM guidelines are non-binding while its 2022 six-hour incident-reporting Directions are binding (both primary); RBI's FREE-AI release is primary, but the report is a set of recommendations, not a direction, and its details still rest on secondary summaries.
What this means for your business
Design for a model disappearing. The June directive took two models offline for every customer for about 19 days, and a Reuters headline later reported OpenAI cutting off its models to a competitor-owned coding tool (H). Route your automation through a provider-neutral layer with a tested fallback model per task, and make any vendor's proposal name a fallback for each critical model. xAI's API is OpenAI-SDK compatible (xAI docs), and DeepSeek's Anthropic-compatible endpoint silently maps any "claude-*" model name to a DeepSeek model, so pin model names explicitly (DeepSeek: Anthropic API).
Budget for monthly migrations. On Haiku 5.5, manual budget_tokens, temperature, top_p, top_k and assistant prefill "can each return a 400 error" (Claude release notes), and products ended on fixed dates, from Grok models on 15 May to Agent Builder on 30 November. Budgets and contracts need a migration allowance, and new builds should avoid tools that already have an end date. The playbook covers delivery practice.
Route by tier and watch introductory prices. Use the bulk tier near $0.10/$0.50 per million tokens for extraction, classification and triage, and the $10/$50 tier only where tests show it changes the outcome. Gemini 3.6, 3.7 and 3.8 Flash double on 1 January 2027 (Google Cloud: generative AI pricing), GPT-5.6 Sol's promotion runs only "at least through" 21 November 2026 (OpenAI pricing), and Gemini 4 Argon goes from $2/$10 to $4/$20 after its introductory period (Google: Gemini 4 Argon). Measure cost per task: the more verbose Grok 4.7 cost $3.74 per Artificial Analysis index task against $1.48 for Grok 4.6 (Artificial Analysis: Grok 4.7). See costs.
Treat data terms as part of model choice. Fable 5.1, Mythos 5.1, Fable 5 and Mythos 5 require 30-day retention, and on any Claude model Batch, the Files API, code execution, the MCP connector, Agent Skills and Managed Agents are not ZDR-eligible (Anthropic, API and data retention). If you need processing in India, the documented options are AWS Bedrock's India profiles and Google Cloud's Mumbai region, both limited to previous-generation models, or self-hosted open weights. xAI keeps API traffic for 30 days by default, and its zero-retention mode disables batch jobs and its retrieval store (xAI: security FAQ). Self-hosted MIT or Apache 2.0 weights avoid the data-location problem, though not agent hijacking.
Build agents to least privilege from day one. OWASP's LLM06 mitigations make a build checklist: narrow tools, read-only identities, authorisation enforced downstream, human approval for high-impact actions and rate limits (OWASP LLM06). Anthropic warns that its computer-use model "will follow commands found in content even when they conflict with your instructions" (Claude: computer use), and the August auto-mode bypass shows why sandboxes, egress allowlists and secret-free prompts are baseline. Controls are in governance.
Benchmark outcome prices and know who sells deployment. Fin, now part of Salesforce, charges $0.99 per outcome with a 50-outcome monthly minimum (Fin pricing), and Zendesk $1.50-2.00 per automated resolution (regional pages differ: £1.30 committed in India and the UK) (Zendesk pricing); compare any per-outcome quote with these. Large buyers can now get deployment help from a lab itself (OpenAI's DeployCo) or from IT firms with lab deals (Tata/TCS, Cognizant, Accenture and HCLTech). The compliance work created by DPDP, the IT Rules and US state laws has fixed dates, listed below.
Weigh vendor-continuity risk explicitly. In this period a loss-making AI segment sat inside a newly listed company whose shares halved from their peak, a model vendor was designated a supply-chain risk, and model access was withdrawn twice, once by directive and once in a reported commercial cut-off. Before you commit, ask each vendor or partner which model IDs are pinned, which fallback has been tested, where data is processed and retained, and which of the deadlines below the build depends on.
Deadlines already on the calendar
| Date | What happens | Conf. | Source |
|---|---|---|---|
| 23 Oct 2026 | Gemini deep-research-pro-preview-12-2025 shuts down; OpenAI legacy snapshots shut down, including o1, o3-mini, o4-mini, gpt-4.1-nano, gpt-4-turbo and gpt-image-1 |
P | Gemini API changelog; OpenAI deprecations |
| 26 Oct 2026 | Colorado AG hearing on its ADMT and Chatbot Safety rules; written comments close (later if the hearing continues) | P | Colorado AG: AI rulemaking |
| 31 Oct 2026 | OpenAI Evals become read-only | P | OpenAI deprecations |
| 12 Nov 2026 | Anthropic 2026 Usage Policy takes effect; EuroHPC AI Gigafactories call closes (trade press; Commission page not reached) | P; S | Anthropic: 2026 usage policy update; EuroHPC: gigafactories call |
| 13 Nov 2026 | DPDP consent-manager provisions (s.6(9), s.27(1)(d)) and Rule 4 commence | P | Gazette G.S.R. 843(E) |
| 21 Nov 2026 | Earliest end of GPT-5.6 Sol promotional pricing | P | OpenAI pricing |
| ~28 Nov 2026 | EO 14434: the APST's proposed federal "Super Intelligence" definition due | P | Federal Register: EO 14434 |
| 30 Nov 2026 | OpenAI Agent Builder, Evals dashboard and API, and reusable prompts shut down; Claude Sonnet 4.5 retires on the Claude API | P | OpenAI deprecations; Claude release notes |
| Dec 2026 | EU prohibition on nudification and CSAM generators applies (P); the watermarking grace period for systems already on the market ends 2 Dec (S) | P; S | EU Commission: AI framework; Addleshaw Goddard: Omnibus |
| 11 Dec 2026 | Older GPT-5 snapshots and o3 removed from the OpenAI API | P | OpenAI deprecations |
| 1 Jan 2027 | Gemini 3.6, 3.7 and 3.8 Flash move to $1.50/$7.50 (3.8 Flash cached input $0.075 to $0.15); Colorado's ADMT law (SB 26-189) and Chatbot Safety Act (HB 26-1263) take effect (the X.AI v. Weiser stay bars AG enforcement of SB 24-205 and of replacing legislation enacted that session; whether it reaches SB 26-189 is unconfirmed); New York's RAISE Act takes effect (amendments reported signed 27 Mar 2026); businesses already using ADMT for significant decisions must meet California's CPPA rules; South Korea's fine-free grace of at least one year ends (secondary); Cal OES's first annual anonymised report on SB 53 frontier-model incidents is due (developer duties have applied since 1 Jan 2026) | P; S; W | Google Cloud: generative AI pricing; Colorado AG: AI rulemaking; CourtListener: X.AI v. Weiser docket; Wikipedia: RAISE Act; CPPA regulations; Presencis: Korea AI Basic Act timeline; Goodwin: SB 53 |
| 6 Jan 2027 | OpenAI accepts no new fine-tuning jobs from anyone; fine-tuning has been closed to new organisations since 7 May 2026 | P | OpenAI deprecations |
| Mar 2027 | GSA OneGov offer of Grok 4 and Grok 4 Fast at $0.42 per agency for 18 months ends ("valid until March 2027"; Wikipedia gives 20 Mar) | P | GSA, 25 Sep 2025 |
| 1 Apr 2027 | OpenAI's gpt-5.3-codex, gpt-5.4-nano and gpt-5.1 shut down | P | OpenAI deprecations |
| First half of 2027 | Nvidia-Hugging Face purchase expected to close, subject to regulatory approvals | P | Nvidia 8-K |
| 13 May 2027 | DPDP core duties commence: notice, consent, security, breach reporting, data-principal rights, cross-border transfer, together with the penalty provisions | P | Gazette G.S.R. 843(E) |
| Sep-Oct 2027 | Earliest retirement dates: Fable 5.1 (1 Sep), Opus 5.5 (22 Sep), Sonnet 5.5 (28 Sep), Haiku 5.5 (7 Oct) | P | Claude: models overview |
| 2 Dec 2027 | EU Annex III high-risk duties (for example hiring and credit scoring) | P | EU Commission: AI framework |
| 2 Aug 2028 | EU Annex I (product-embedded) high-risk duties | P | EU Commission: AI framework |
Open questions and unverified items
The conflicts below appear in the sources and are shown rather than silently resolved; the "Treatment" column says which date or figure this report uses. Date conflicts already shown inside the table rows (Sora, GPT-6 Astra, Ofcom, Indonesia's Grok block, ChatGPT ads, Grok Build, GLM-5, Qwen3.5, DeepSeek V4-Pro GA, Manus, the summit's end date, SB 53's signing date) are not repeated here.
| Item | What the sources say | Treatment here |
|---|---|---|
| Anthropic $65B run-rate (July 2026) | CNBC, Bloomberg and Reuters headlines only (HN: Bloomberg on Anthropic run-rate); Anthropic's latest own figure is above $47B (May) | Unconfirmed by Anthropic |
| Kimi K3 weights | Hugging Face repo created 13 Jun; OpenRouter listing and launch 16 Jul; one earlier read dated the HF listing 27 Jul; Microsoft's Q2 2026 report says the full weights were released in July (Microsoft AI Diffusion Report 2026 Q2) | Launch 16 Jul; weights July 2026 |
| Bartz settlement | About $3,000 per title (Wikipedia 2025 page) vs about $3,100 per work and final approval on 20 Jul 2026 (earlier read, not re-verified) | $1.5B total only |
| Microsoft's India commitment | $50B by 2030 for lower-income countries (Wikipedia, summit); $17.5B (Fortune); INR 25,000 crore for 2025-27 (Wikipedia, AI in India) | Not reconciled |
| Summit investment total | Over $200B "expected" over two years (Fortune); an earlier "$250B" summary is unsupported | $200B, as an expectation |
| Colorado | SB 26-189 confirmed on the legislature's page; the US intervention (24 Apr 2026) and the 27 Apr minute order (ECF 24) are docket-verified (primary) (CourtListener: X.AI v. Weiser docket); no docket entries after 27 Apr were read | Stay shown in April 2026; whether it reaches SB 26-189 is arguable and unconfirmed |
| New York RAISE amendments | Reported passed 11 Mar and signed 27 Mar 2026 (secondary summaries); Wikipedia: RAISE Act (cached Feb 2026) still shows them in committee; nysenate.gov and governor.ny.gov blocked | Shown in March 2026 as reported; conflict kept visible |
| Pentagon and Anthropic | Designation directed 27 Feb (H) and confirmed by BBC as February; BBC says use stopped "late August," with Claude still used in operations "as recently as last week" before 5 Oct | Both shown |
| Grok 4.6 benchmark | Artificial Analysis index 44 (P) vs a secondary "61," which matches the model's output speed | 44 |
| AWS Bedrock India In-Region listing | The Regional-availability page lists Claude Opus 5 and Sonnet 5 as In-Region in Mumbai on the bedrock-mantle endpoint, but the two model cards' own bedrock-mantle tables omit Mumbai (AWS Bedrock regional availability) |
India cross-region (in.) profiles used; single-region Mumbai to be confirmed in the console |
| Texas TRAIGA effective date | "1 Jan 2026" comes from earlier secondary sources; it was not found in the enrolled text, which confirms penalties of $80,000-200,000 and up to $40,000 a day (Texas Legislature: HB 149 enrolled) | 1 Jan 2026 kept, labelled Q |
| US executive-order follow-through | No EO 14409 implementation output, and none of the EO 14365 Commerce list, BEAD notice, FCC proceeding or FTC statement, was found; DOJ suits against California, Utah or Texas rest on search summaries only | Orders shown; outputs and suits left out |
| NIST CAISI to CAISSI | NIST's page is now titled "Center for Advancing Innovation and Standards for Super Intelligence (CAISSI)"; evaluation posts still say CAISI (NIST: CAISSI) | Rename date not established; its link to EO 14434 is an inference |
| Codex weekly users | About 2M (March, Wikipedia); more than 5M (2 Jun, vendor figure via press); about 10M combined with ChatGPT Work (21 Jul, secondary, a combined metric); no OpenAI page read | March and June figures shown with labels |
| TRAI TCCCPR Third Amendment | Notification on 18 Sep 2026 rests on secondary reports; the final text was not read, and no explicit AI-voice rule is confirmed | Shown as reported, tag S |
| Undated rounds | Harvey raised $550M at $15.5B (Harvey) and Clay $115M at $7.1B (Clay), but neither page shows a date | Excluded from monthly tables |
| Contested science claims | OpenAI's claimed Cycle Double Cover proof on 10 Jul 2026 (OpenAI: CDC proof PDF, H), its Navier-Stokes claim (8 Sep) and 377 open problems (6 Oct), and a letter from 28 Fields Medallists warning against using open problems as benchmarks (11 Sep) (Wikipedia: 2026 in AI, W) | Left out of the monthly tables; not independently assessed |
Several gaps remain. This research set has no verified 2026 status for Nvidia H200 sales to China. No first-party API processes requests in India (confirmed, primary), and AWS Bedrock's India profiles cover only previous-generation models (see the India patterns above). There is no published action-taken report or MeitY enforcement decision on the Grok notice (government sources reported about 3,500 items blocked and 600 accounts deleted around 11 Jan). Indian AI funding rounds beyond Sarvam AI are missing, as are OpenAI model releases between August 2025 and June 2026; A2A's 2026 status rests on the project's own blog. Independent benchmark figures are now partly extracted (the Terminal-Bench 4.0 board, LMArena's 8 Oct snapshot and the Artificial Analysis index; see the Models thread), but the official SWE-bench Verified board has no 2026 frontier rows, and the 2026 ARC-AGI-2 values exist only in Epoch's copy with a blank provenance column (single-source; kept out of this timeline). Wikipedia's 2026 page is itself flagged for verification, so every W item should be checked against the cited outlet before it is relied on.
Sources
All sources were read for this research up to 12 October 2026. Entries marked "re-read 11 Oct 2026" were checked live on 11 October; entries marked "read 12 Oct 2026" were added on 12 October. Confidence tags in the tables show how each source was used.
- Anthropic: Fable and Mythos access (re-read 11 Oct 2026)
- Washington Post: vetting
- Google: Gemini 4 Argon (re-read 11 Oct 2026)
- Anthropic: Cyber Verification Program
- OpenAI: GPT-6 Astra system card (re-read 11 Oct 2026)
- BBC: GPT-6.1 and agent incidents (re-read 11 Oct 2026)
- Claude release notes (re-read 11 Oct 2026)
- Anthropic: Haiku 5.5
- OpenAI pricing
- DeepSeek pricing
- MCP blog
- TechCrunch: Dots
- ABC: AI agent accessed government site
- Simon Willison: prompt injection (re-read 11 Oct 2026)
- Anthropic: draft S-1
- SpaceX 424B4 (re-read 11 Oct 2026)
- SpaceX 8-K (re-read 11 Oct 2026)
- TechCrunch: OpenAI chip
- Alphabet 10-Q
- EU Commission: AI framework (re-read 11 Oct 2026)
- BBC: Pentagon and Anthropic
- Gazette G.S.R. 843(E)
- DeepSeek release list (re-read 11 Oct 2026)
- Wikipedia: Stargate LLC
- Wikipedia: OpenAI
- Federal Register: EO 14179 (re-checked via the Federal Register API, 11 Oct 2026)
- Wikipedia: 2025 in AI
- Introl: DeepSeek bans
- Hugging Face: Anthropic Economic Index
- SiliconANGLE: France EUR 109B
- Wikipedia: Grok (chatbot)
- Wikipedia: Lovable
- Anthropic: Series E
- Wikipedia: Manus
- GitHub API: openai-agents-python
- Wikipedia: Cognition AI
- Wikipedia: Moveworks
- n8n press
- Wikipedia: Workato
- Wikipedia: 01.AI
- Wikipedia: SpaceXAI
- Invariant: tool poisoning
- GitHub API: adk-python
- Wikipedia: OpenAI Codex
- Wikipedia: Sarvam AI
- Anthropic: Series G
- Invariant: GitHub MCP
- Wikipedia: Telegram
- GitHub API: claude-agent-sdk-python
- The Verge: Meta and Scale
- Wikipedia: AI in India
- Nelson Mullins: TRAIGA
- Hugging Face API: Baidu
- Wiz: malware invoking AI
- The Register: Replit
- Federal Register API: 2025 AI orders (re-read 11 Oct 2026)
- HN: Coze Studio
- Wikipedia: Nvidia
- Wikipedia: Visakhapatnam
- Hugging Face: gpt-oss-120b
- Storyboard18: RBI FREE-AI
- HN: ChatGPT Go India
- Qwen blog
- Brave: Comet prompt injection
- Cloudflare: Salesloft Drift
- BleepingComputer: s1ngularity
- Wikipedia: Regulation of AI
- Greenberg Traurig: Colorado
- Wikipedia: Workday
- Mistral news
- Wikipedia: Anthropic
- The Verge: Statsig
- Google Cloud: AP2
- Workday: Sana
- HN: Kimi OK Computer
- Clifford Chance: SB 53
- NIST CAISI: DeepSeek evaluation
- Anthropic: Google Cloud TPUs
- LangChain blog
- Wikipedia: ChatGPT Atlas
- Brave: unseeable prompt injections
- OpenAI: Sky acquisition
- Analytics Vidhya: Jio and Google AI Pro
- Business Today: ChatGPT Go free
- Wikipedia: Stability AI
- Indian Infrastructure: AI governance guidelines
- MeitY: DPDP Rules
- Anthropic: disrupting AI espionage
- Wikipedia: Anysphere
- Salesforce: Informatica
- Workday: Pipedream
- Federal Register: EO 14363 (re-checked via the Federal Register API, 11 Oct 2026)
- Yahoo Finance: Meta and Google TPU talks
- Wikipedia: Automation Anywhere
- TechCrunch: OpenAI and Tata
- Anthropic: Bun
- Salesforce Q3 FY26
- NIST CAISI
- Anthropic: MCP and AAIF (re-read 11 Oct 2026)
- OWASP: Agentic Top 10
- Federal Register: EO 14365 (re-checked via the Federal Register API, 11 Oct 2026)
- Paul Hastings: preemption EO
- Wikipedia: RAISE Act
- Wikipedia: Groq
- Anthropic: Genesis Mission
- Wikipedia: Grok deepfake scandal
- Wikipedia: TRAIGA
- exchange4media: MeitY notice
- Wikipedia: 2026 in AI
- Wikipedia: Zhipu AI
- Wikipedia: MiniMax
- Weintraub: task force
- Anthropic: Economic Index Jan 2026
- Wikipedia: ChatGPT
- Anadolu: EU DSA probe
- StorageReview: Maia 200
- Kimi K2.5 blog
- Wikipedia: Decagon
- Wikipedia: Cerebras
- Wikipedia: ServiceNow
- Wikipedia: Colossus
- SpaceX: xAI joins SpaceX
- Alphabet Q4 2025 release
- ElevenLabs: Series D
- Axios: Opus 4.6
- Business Today: IT Rules
- Hugging Face API: zai-org
- Wikipedia: Doubao
- Anthropic: Bengaluru office
- Anthropic: India brief
- Hugging Face API: Qwen
- Wikipedia: AI Impact Summit
- Gemini API changelog
- DeepMind: Gemini 3.1 Pro
- Google: Deep Think
- Fortune: summit closes
- Salesforce Q4 FY26
- Wikipedia: StepFun
- Anthropic: Vercept
- VentureBeat: Qwen3.5
- X post: Secretary of War
- TechCrunch: OpenAI $110B
- SpaceX S-1/A #2
- Wikipedia: SambaNova
- TechNode: DeepSeek V4
- HN: Bloomberg $20B
- BBC: India Supreme Court
- WSJ: formal label
- CNBC: Oracle
- Axios: Moltbook
- Wikipedia: Legora
- Hollywood Reporter: Sora
- Anthropic: Economic Index Mar 2026
- CNN: injunction
- CNBC: $852B round
- OpenRouter models
- DeepMind: Gemma 4
- OpenAI: TBPN
- Anthropic: Google and Broadcom
- Anthropic: Mythos Preview system card
- Meta: Muse Spark
- HN: Stargate 2026
- Hugging Face API: Tencent
- Anthropic: Economic Index Jun 2026
- Google: TPU 8
- Semafor: 75% of code
- CNBC: Manus
- Google: AP2 to FIDO
- Economic Times: Vizag
- Alphabet Q1 2026 release
- Wikipedia: Ernie Bot
- Bloomberg Law: 10 GW
- AI Weekly: capex
- xAI docs: Grok 4.3
- NIST CAISI: V4-Pro
- Wikipedia: SpaceX IPO
- ElevenLabs blog
- Colorado General Assembly: SB26-189 (re-read 11 Oct 2026)
- xAI: 15 May retirement
- Anthropic: Stainless
- Qwen 3.7 post
- Google: Gemini 3.5
- WSJ: OpenAI IPO
- Hugging Face API: Step-3.7-Flash
- Salesforce Q1 FY27
- Anthropic: Series H (re-read 11 Oct 2026)
- MiniMax M3 licence
- BBC: Gemini security test
- Anthropic: Mythos
- OpenAI deprecations
- CNBC: Google and SpaceX
- MacRumors: Apple AI
- Anthropic: Fable 5 and Mythos 5
- The Verge: Fable guardrails (URL truncated in the Hacker News listing; not a working link, and no full URL was found)
- Hugging Face: Kimi K2.7-Code
- Salesforce: Fin agreement
- Pew Research: chatbot use
- MCP: enterprise-managed auth
- Google: computer use in 3.5 Flash
- OpenAI Deployment Safety Hub (re-read 11 Oct 2026)
- Semafor: Mythos
- Al Jazeera: restrictions lifted
- Alphabet Q2 2026 release
- Addleshaw Goddard: Omnibus
- Hugging Face API: Hy3
- xAI: Grok 4.5
- OpenAI: GPT-5.6
- Google: Agent Platform pricing
- GitHub: grok-build
- Kimi K3 blog
- Hugging Face: Kimi-K3
- Qwen 3.8 post
- Kimi post
- Google: Gemini 3.6 Flash
- Anthropic: Opus 5
- CNBC: open weights
- NatLawReview: Omnibus in force
- MCP roadmap
- MiniMax-H3 README
- Microsoft FY26 Q4
- EuroHPC: gigafactories call
- Hugging Face: Qwen3.8-27B
- Qwen3.8-Max licence
- Anthropic: Fable
- The Register: Taalas (URL as given in the Hacker News listing returned 404; not a working link, and no corrected URL was found)
- xAI: Grok Bot
- xAI: Grok 4.6
- Lovable blog
- DeepSeek Harness
- Google: Gemini 3.7 Flash (URL truncated in the Hacker News listing; not a working link, and no full URL was found)
- ainauten: DGX Spark price
- Bloomberg Law: Cursor
- HN: CNBC on Anthropic run-rate
- HN: Reuters on Anthropic run-rate
- Claude: Agent Skills
- DeepSeek updates
- Salesforce Q2 FY27
- Salesforce: Claudeforce
- OpenAI changelog
- DuckLabs
- Bloomberg: Ox Alpha
- Business Insider: Nvidia and Hugging Face
- Hugging Face API: Hy4-preview
- HN: SpaceX Cursor
- Anthropic: Enterprise Frontier Safeguards
- DeepMind: Gemini 3.8 Flash model card
- Google Cloud: generative AI pricing
- The Verge: GPT-6 Astra
- CNBC: Nvidia and Hugging Face
- HN: Huawei Ascend 950
- Wikipedia: Opposition to AI data centers
- Cognition: Series E
- Salesforce: Fin completed
- Forkast: Cohere
- Factory news
- Anthropic newsroom
- Wikipedia: NYT v. Microsoft and OpenAI
- xAI: Grok 4.7
- Artificial Analysis: Grok 4.7
- Guardian: British Columbia suit
- Anthropic: Opus 5.5
- METR: Opus 5.5
- OpenAI: GPT-6 Sol and Luna
- CNBC: appeals court
- Anthropic: Sonnet 5.5
- ElevenLabs: tender
- Anthropic: 2026 usage policy update
- OpenAI: GPT-6 October update
- OpenAI: GPT-6 for everyone
- Anthropic: Cyber Mission
- Deno: Cloudflare
- OpenRouter rankings
- Kimi K3 licence
- Claude pricing
- npm: Claude Code downloads
- Stack Overflow 2025
- Epoch AI: Stargate sites
- Yahoo Finance: hyperscaler capex
- MobileTime: Brazil AI bill
- arXiv: EchoLeak
- Gazette: DPDP Act
- Outlook Business: OpenAI offices
- IANS: data-centre capacity
- Claude: data residency
- xAI docs
- DeepSeek: Anthropic API
- xAI: security FAQ
- OWASP LLM06
- Claude: computer use
- Fin pricing
- Zendesk pricing
- Colorado AG: AI rulemaking
- Claude: models overview
- HN: Bloomberg on Anthropic run-rate
- Harvey
- Clay
- OpenAI: CDC proof PDF
- Hogan Lovells: IT Rules (read 11 Oct 2026)
- OpenAI: data controls (read 11 Oct 2026)
- xAI: regional endpoints (read 11 Oct 2026)
- Goodwin: SB 53 (read 11 Oct 2026)
- Microsoft AI Diffusion Report 2026 Q2
- Federal Register (read 12 Oct 2026)
- ServiceNow release, SEC (read 12 Oct 2026)
- CoreWeave 424B4 (read 12 Oct 2026)
- CoreWeave 8-K, 25 Sep 2025 (read 12 Oct 2026)
- Meta 10-Q (read 12 Oct 2026)
- CERT-In guidelines v2.0 (read 12 Oct 2026)
- RBI press release (read 12 Oct 2026)
- Linux Foundation: AAIF formation (read 12 Oct 2026)
- Business Today: MeitY notice (read 12 Oct 2026)
- Storyboard18: X takedowns (read 12 Oct 2026)
- legislation.gov.uk: DUAA s.80 (read 12 Oct 2026)
- iPleaders: IT Rules 2026 (read 12 Oct 2026)
- Gala Law: IT Rules (read 12 Oct 2026)
- Helicone: joining Mintlify (read 12 Oct 2026)
- Promptfoo blog (read 12 Oct 2026)
- Secure Privacy: NY AI law summary (read 12 Oct 2026)
- Bloomberg Law: $852B round (read 12 Oct 2026)
- GitHub releases: Agent Framework (read 12 Oct 2026)
- CourtListener: X.AI v. Weiser docket (read 12 Oct 2026)
- TechRadar: Stargate UK halted (read 12 Oct 2026)
- Daily Maverick: policy withdrawn (read 12 Oct 2026)
- Constellation Research: Codex users (read 12 Oct 2026)
- GOV.UK: Andy Burnham (read 12 Oct 2026)
- GOV.UK: Ministerial appointments, July 2026 (read 12 Oct 2026)
- Flowise discussion #6727 (read 12 Oct 2026)
- Nvidia 8-K (read 12 Oct 2026)
- Federal Register: EO 14434 (read 12 Oct 2026)
- NIST: CAISSI (read 12 Oct 2026)
- Accenture Q4 FY26 8-K (read 12 Oct 2026)
- Outlook Business: TCS Q2 (read 12 Oct 2026)
- LMArena: text (read 12 Oct 2026)
- LMArena: WebDev (read 12 Oct 2026)
- Terminal-Bench (read 12 Oct 2026)
- Artificial Analysis (read 12 Oct 2026)
- Câmara dos Deputados: PL 2338/2023 (read 12 Oct 2026)
- CERT-In Directions 2022 (read 12 Oct 2026)
- Elets eGov: IndiaAI charge (read 12 Oct 2026)
- CareerIndia: TRAI rules (read 12 Oct 2026)
- TRAI-hosted comments on the draft (read 12 Oct 2026)
- IndiaAI price calculator (read 12 Oct 2026)
- IndiaAI empanelled GPUs (read 12 Oct 2026)
- AWS Bedrock regional availability (read 12 Oct 2026)
- Google Cloud: data residency (read 12 Oct 2026)
- Anthropic, API and data retention (read 12 Oct 2026)
- CPPA regulations (read 12 Oct 2026)
- Presencis: Korea AI Basic Act timeline (read 12 Oct 2026)
- GSA, 25 Sep 2025 (read 12 Oct 2026)
- Texas Legislature: HB 149 enrolled (read 12 Oct 2026)