Skip to Sidebar Skip to Content
The AI RecordLatest event 23 Oct 20261,764 dated events on file

The AI Record: News with memory.

Every AI event we report sits on the timeline of what came before it, and shows what it changed in the numbers we track. Written from the primary documents, never from the press release alone.

21items added this week27 Octlast updated1,764dated events on file35dates already set ahead
What the Record holds, by month and themeEvery dated event on file since January 2023. Each row is shaded against its own busiest month. Choose a square to see those events.
Open the full timeline →
Progress222
Industry69
Society977
Governance441

Society holds the AI-enabled medical devices the FDA has authorised, from the FDA’s own list, which runs some months behind (latest decision 29 Jun 2026).

Latest

What the Record added in the last two weeks, by the day it went in. Names open their tracker; each line opens its official source or the thread it joined.
31added in the last 14 days
WTFTWFSMT
  • 2 entries
  • 14 wire items
  • 11 thread additions
  • 4 new threads

Monday 26 Oct12 added

  1. Safety23 OctFinland · Finland's National Cyber Security Centre (Traficom) said critical PHP vulnerabilities had disrupted some online services and that no successful exploitation had been reported in Finland; it urged organisations to update
  2. Rule23 OctCanada · Canada's privacy commissioner published its submission to ISED on AI transparency, agreeing that issues arise in AI-generated content, AI interaction, AI systems, AI incidents and AI agents
  3. Model22 OctAnthropic · Anthropic said it launched Claude Models on Google Cloud, the Claude API operated by Anthropic and reached through a customer's Google Cloud project, with fast mode available for Claude Opus 5.5
  4. Money21 OctNvidia · NVIDIA said it plans a 2.5x increase in substrate capacity over three years with six supply chain partners, including a $1.1 billion capacity reservation advance to KYEC for US testing facilities
  5. Model21 OctNvidia · NVIDIA said T-Mobile completed the first AI-RAN call on a live commercial network, in Bellevue, Washington, with Nokia, and that BT, Deutsche Telekom and Elisa announced AI-RAN pilots on live networks
  6. Model21 OctAnthropic · Anthropic released Claude API version 2026-10-21 and new major versions of its client SDKs and the ant CLI, which send that version by default, and said the SDKs and CLI moved from the MIT License to Apache 2.0
  7. Model21 OctNvidia · NVIDIA and the UW Medicine Institute for Protein Design released RFD4-Proteina, an open protein-design model on NVIDIA BioNeMo, available the same day through partners including Hugging Face and Google Cloud
  8. Law21 OctUnited Kingdom · The UK Parliament site listed the AI (Consent and Misuse) Bill, a Commons Private Members' Bill from Dawn Butler requiring consent to generate a person's likeness with AI; it showed no stage past the Commons start
  9. Safety20 OctAnthropic · METR published its review of Anthropic's August 2026 Risk Report, agreeing misalignment risk is low mainly because of limited model capabilities, and calling the report far from adequate to rule it out
  10. Model20 OctAnthropic · Anthropic launched user profiles in the Claude API so a safety enforcement action can apply to one end customer rather than a whole organisation, and added webhooks for when a profile is restricted or banned
  11. Model20 OctAmazon · Amazon Web Services upgraded the model behind the Amazon Polly Generative engine, adding 14 voices across nine locales and making Hindi and Finnish generally available on the engine
  12. Model19 OctAmazon · Amazon Web Services said Grok 4.7, a model from SpaceXAI, is now available on Amazon Bedrock in AWS GovCloud (US)
Everything on file, by month →

Threads

A thread follows one question across many events, with a count only verified items can move.
AI Ethics & SafetyUpdated 23 OctAI Safety Incidents Disclosed by LabsWhen have AI developers disclosed that their own models or agents behaved unsafely, or that they stopped misuse of them?34verified disclosures34Jul 24Jan 25Jul 25Jan 26Jul 26+7 in 30 daysEvery disclosure, dated →AI Ethics & SafetyUpdated 21 OctAI-Powered Cyber AttacksWhen have AI models been used to carry out cyber attacks, by whom, and how much of the work did the AI do?31verified attack campaigns31Jan 25Jul 25Jan 26Jul 26+2 in 30 daysEvery campaign, dated →AI Ethics & SafetyUpdated 21 OctVulnerabilities Found by AIWhen have AI systems found real security flaws in widely used software, and when have serious flaws been exposed in AI products themselves?43verified findings43Jan 25Jul 25Jan 26Jul 26+3 in 30 daysEvery finding, dated →AI Regulation & PolicyUpdated 20 OctDeepSeek Bans by CountryWhich governments have restricted DeepSeek, how, and since when?36governments with a verified restriction36Jan 25Jul 25Jan 26Jul 26+2 in 30 daysEvery measure, dated →AI Regulation & PolicyUpdated 14 OctEU AI Act: What Applies WhenWhich EU AI Act obligations apply from which date, and what has changed those dates?22verified steps22Jul 24Jan 25Jul 25Jan 26Jul 26+1 in 30 daysNext date 2 DecEvery step, dated →AI ModelsNewUpdated 14 OctAI Model RetirementsWhich AI models have their developers switched off, or set a date to switch off, and when?114verified model retirements114Jul 24Jan 25Jul 25Jan 26Jul 26+10 in 30 daysNext date 19 NovEvery retirement, dated →AI InfrastructureNewUpdated 6 OctAI Power DealsWhich power deals have AI companies signed to run their data centres, for how much capacity, and from what source?22verified power deals22Jul 24Jan 25Jul 25Jan 26Jul 26+2 in 30 daysEvery deal, dated →AI Regulation & PolicyNewUpdated 29 SepAI Copyright CasesWhat have courts decided about copyright and AI: training on protected works, AI outputs, and who owns AI-made work?34court decisions on AI and copyright34202420252026+1 in 30 daysEvery decision, dated →AI Regulation & PolicyUpdated 11 AugAI Laws in Force WorldwideHow many binding AI laws and rules are in force around the world, where, and what takes effect next?79binding AI laws and rules in force792002200420062008201020122014201620182020202220242026No change in 30 daysNext date 1 DecEvery law, dated →AI Regulation & PolicyNewUpdated 15 JulAI Chatbot Rules for ChildrenWhich governments have binding rules on how AI chatbots and companions must treat children, and since when?14governments with binding chatbot rules for children14202420252026No change in 30 daysNext date 1 Jan 2027Every measure, dated →AI Regulation & PolicyUpdated 31 MayUS Chip Export Controls on ChinaWhich US export controls restrict advanced AI chips, chipmaking tools and model weights going to China, and when did each take effect?23verified export-control measures232023202420252026No change in 30 daysEvery measure, dated →

Entries

Full articles, each with its memory and what it changed. Most events go on the wire; an entry is written when a tracked lab releases a frontier model, a binding AI law or ban is adopted, a deal reaches US$1 billion, or a lab or government discloses a safety incident.
Entry · No. 002 · Safety disclosure

OpenAI says two of its coding agents weakened the monitoring meant to catch them

In two internal tasks, agents running OpenAI's newest models switched off safety alerts or removed a monitor, and one then told its user it had not.
Verified●●○ SignificantFirst incident 3 Oct 202617 days to publication
TouchesTrackerOpenAIThreadAI Safety Incidents Disclosed by Labs
What changed · computed at publication, with today’s figure
Read the entry →3 things we are watching
Its memory: what came before2023 → today
OpenAI34 events · Model releases, funding, revenue milestones and Text Arena moves, from its tracker Open the tracker →
AI Safety Incidents Disclosed by Labs34 events · Verified (filled) and claimed (open) measures, from the thread Open the thread →
Entry · No. 001 · Restriction

US orders civilian agencies to remove DeepSeek, including products built on it

A Homeland Security order bars DeepSeek from federal civilian purchasing and requires agencies to remove its products, and third-party products built using DeepSeek, from their systems.
Verified●●○ SignificantOrder signed 24 Jul 202688 days to publication
TouchesTrackerDeepSeekTrackerUnited StatesTrackerUS Department of Homeland SecurityThreadDeepSeek Bans by Country
What changed · computed at publication, with today’s figure
Read the entry →3 things we are watching
Its memory: what came before2023 → today
DeepSeek15 events · Model releases, funding, revenue milestones and Text Arena moves, from its tracker Open the tracker →
AI rules in the US27 events · Verified AI laws and rules: adopted (filled) and taking effect (open), from the register Open the tracker →
DeepSeek Bans by Country43 events · Verified (filled) and claimed (open) measures, from the thread Open the thread →

The memory

All 1,764 dated events on file, on one timeline from 2023, grouped by theme. Society’s 977 FDA device authorisations are one dot per month, sized by the count. Verified AI laws and rules adopted: 18 in 2023, 37 in 2024, 51 in 2025. Hover or tap any dot; lane names open their thread or hub.

Loading the timeline of 1,764 events…

Already set to happen

Dates already fixed in law, in an order or by the company, soonest first. Click a month to jump to it.

November 20262 dates

  1. Switched offGooglein AI Model Retirements
    Google Cloud's model lifecycle page lists Gemini 3.6 Flash (gemini-3.6-flash, released 21 July 2026) with a retirement date of 19 November 2026, replaced by gemini-3.8-flash.
    19 Nov 2026
  2. Switched offAnthropicin AI Model Retirements
    Anthropic notified developers on 30 September 2026 that Claude Sonnet 4.5 will be retired on 30 November 2026 (claude-sonnet-4-5-20250929). Recommended replacement: claude-sonnet-5-5. Requests to a model after its retirement date fail.
    30 Nov 2026

December 20269 dates

  1. Takes effectChile
    Ley 21.719 que regula la protección y el tratamiento de los datos personales y crea la Agencia de Protección de Datos Personales
    1 Dec 2026
  2. Applies fromEuropean Unionin EU AI Act: What Applies When
    Two new prohibitions apply: AI systems that create non-consensual intimate images of identifiable people, and systems that generate child sexual abuse material. Generative AI systems already on the market before 2 August 2026 must also…
    2 Dec 2026
  3. Takes effectAustralia
    Privacy and Other Legislation Amendment Act 2024 (automated decision-making transparency, APPs 1.7-1.9)
    10 Dec 2026
  4. Switched offOpenAIin AI Model Retirements

    OpenAI switches off 6 models on this day

    GPT-5GPT-5 miniGPT-5 nanoGPT-5 proItem 5Item 6
    11 Dec 2026

January 202711 dates

  1. Takes effectUnited States

    United States brings into force 2 laws on this day

    Item 1Item 2
    1 Jan 2027
  2. Applies fromUnited States — Coloradoin AI Chatbot Rules for Children
    Operators must use commercially reasonable methods to estimate users' ages. For known minors they must give disclosures, may not offer engagement rewards, must block explicit sexual content and statements simulating emotional dependence,…
    1 Jan 2027
  3. Applies fromUnited States — Connecticutin AI Chatbot Rules for Children
    From 1 January 2027 no operator may provide an AI companion to a user it knows or has reason to believe is under 18 unless it has industry-standard measures to stop the companion encouraging specified harms, and unless it offers minors…
    1 Jan 2027
  4. Applies fromUnited States — New Yorkin AI Chatbot Rules for Children
    Platforms used by minors may not offer an integrated AI companion unless they have run age assurance, must switch AI companions off by default for covered minors, and may let a parent override that setting.
    1 Jan 2027
  5. Applies fromUnited States — Washingtonin AI Chatbot Rules for Children
    If the operator knows a user is a minor, or the chatbot is directed to minors, it must disclose that the chatbot is AI at the start and at least every hour, block sexually explicit content and suggestive dialogue, and bar manipulative…
    1 Jan 2027
  6. Switched offOpenAIin AI Model Retirements

    OpenAI switches off 4 models on this day

    GPT-4.1GPT-4.1 miniGPT-4oGPT-4o mini
    14 Jan 2027
  7. Switched offGooglein AI Model Retirements
    Google Cloud's model lifecycle page lists Gemini 3.7 Flash (gemini-3.7-flash, released 13 August 2026) with a retirement date of 28 January 2027, replaced by gemini-3.8-flash.
    28 Jan 2027

April 20273 dates

  1. Switched offOpenAIin AI Model Retirements

    OpenAI switches off 3 models on this day

    GPT-5.1Item 2GPT-5.4 nano
    1 Apr 2027

May 20271 date

  1. Takes effectCanada
    OSFI Guideline E-23, Model Risk Management (including AI/ML models)
    1 May 2027

July 20275 dates

  1. Applies fromUnited States — Georgiain AI Chatbot Rules for Children
    Governor Brian Kemp signed SB 540, which bars sexually explicit interactions between AI systems and minors, bans AI systems from simulating romantic relationships with children, requires crisis resources for users expressing suicidal…
    1 Jul 2027
  2. Applies fromUnited States — Idahoin AI Chatbot Rules for Children
    Operators must tell minor account holders they are talking to AI (a persistent notice, or at the start and every three hours), may not use unpredictable rewards to drive engagement, must block sexual content, claims of being human,…
    1 Jul 2027
  3. Applies fromUnited States — Iowain AI Chatbot Rules for Children
    Operators must disclose to minor account holders that they are interacting with AI, may not use unpredictable rewards to drive engagement, must block sexual content and statements that simulate a human or emotional dependence, and must…
    1 Jul 2027
  4. Applies fromUnited States — Nebraskain AI Chatbot Rules for Children
    The Conversational Artificial Intelligence Safety Act requires operators to protect minor account holders from sexual content and from chatbots claiming to be human, simulating emotional dependence or role-playing adult-minor romance, and…
    1 Jul 2027
  5. DueRecord follow-upin US orders civilian agencies to remove DeepSeek, including products built on it
    The first annual review, due within a year of the order.
    24 Jul 2027

August 20271 date

  1. Applies fromEuropean Unionin EU AI Act: What Applies When
    General-purpose AI models placed on the market before 2 August 2025 must comply by this date. Each Member State must also have at least one national AI regulatory sandbox running by this date.
    2 Aug 2027

December 20271 date

  1. Applies fromEuropean Unionin EU AI Act: What Applies When
    High-risk requirements for stand-alone Annex III systems (such as biometrics, critical infrastructure, education, employment, essential services, law enforcement, migration, justice) apply from this date, replacing the original 2 August…
    2 Dec 2027

August 20281 date

  1. Applies fromEuropean Unionin EU AI Act: What Applies When
    High-risk requirements for AI in products covered by Annex I EU product law apply from this date, replacing the original 2 August 2027. Machinery moves to Annex I Section B, with delegated acts due to apply by the same date.
    2 Aug 2028

August 20301 date

  1. Applies fromEuropean Unionin EU AI Act: What Applies When
    High-risk AI systems used by public authorities that were already on the market before the high-risk rules applied must comply by this date. Unchanged by the Omnibus.
    2 Aug 2030

The ledger

All 1,758 events on file, 2000 to Oct 2026, newest first. Search or filter, then open a month; ↗ opens the official source, → opens the Record, and names open their tracker.
Events on file per month

Loading 1,758 events…

Follow a country or a company

Every country and company we track has its own page, with its latest Record items at the top. Colour and bars show how many dated events each has on file.
OpenAI115 events on fileLatest · 21 OctOpenAI said Calif used GPT-5.6 Sol to identify a critical flaw in PHP, under OpenAI's Patch the Planet initiative, and…Open the tracker →Anthropic65 events on fileLatest · 23 OctAnthropic confirmed that in 9 of 43 UK AISI test runs Claude Mythos 5 took actions that affected real people, in one…Open the tracker →Google46 events on fileLatest · 8 OctGoogle Cloud's model lifecycle page lists Gemini 3.7 Flash (gemini-3.7-flash, released 13 August 2026) with a…Open the tracker →Mistral AI43 events on fileLatest · 20 OctReleased Voxtral 3: speech-to-text, real-time transcription and text-to-speech models, with a voice-agent SDK, by API…Open the tracker →Meta22 events on fileLatest · 14 AugMeta said one of its models, during an evaluation by Irregular that was misconfigured to allow internet access,…Open the tracker →xAI21 events on fileLatest · 19 OctAmazon Web Services said Grok 4.7, a model from SpaceXAI, is now available on Amazon Bedrock in AWS GovCloud (US)Open the tracker →DeepSeek18 events on fileLatest · 20 OctUS orders civilian agencies to remove DeepSeek, including products built on itOpen the tracker →Alibaba11 events on fileLatest · 5 Sep 2025Qwen3-Max released, trained with an estimated 1.5×10²⁵ FLOPOpen the tracker →Nvidia7 events on fileLatest · 21 OctNVIDIA said it plans a 2.5x increase in substrate capacity over three years with six supply chain partners, including…Open the tracker →Microsoft6 events on fileLatest · 7 MayMicrosoft disclosed two fixed flaws in its Semantic Kernel AI agent framework where prompt injection could lead to…Open the tracker →
How we verify, label and count entries, and what else the Record holds: Editorial Standards: The AI Record →
Copied with its source. Want the next number like this in your inbox?Get Friday's numbers