Issue archive

The week-by-week record of frontier AI failures.

357 wires across 141 issues, grouped by ISO week of incident capture. Drawn from public regulator releases, provider changelogs, model-comparison probes, and reader tip-offs.

2026

  1. W24 · 08 Jun–14 Jun sev-1 · Model aws-bedrock-claude-sonnet-4.5-eu-west-2 live-probe citation-pubmed-v1 moved -3.67σ f… 5 wires
  2. W23 · 01 Jun–07 Jun sev-1 · Model aws-bedrock-claude-sonnet-4.5-eu-west-2 live-probe citation-pubmed-v1 moved -3.67σ f… 17 wires
  3. W22 · 25 May–31 May sev-1 · Model aws-bedrock-claude-sonnet-4.5-eu-west-2 live-probe citation-pubmed-v1 moved -3.67σ f… 54 wires
  4. W21 · 18 May–24 May sev-1 · Model anthropic-claude-haiku-4.5 live-probe citation-pubmed-v1 moved -3.67σ from rolling m… 15 wires
  5. W20 · 11 May–17 May sev-1 · Model anthropic-claude-haiku-4.5 live-probe citation-pubmed-v1 moved -3.67σ from rolling m… 13 wires
  6. W19 · 04 May–10 May sev-3 · [eu-ai-office] Three studies on technical solutions to mark and detect AI-generated conten… 10 wires
  7. W18 · 27 Apr–03 May sev-3 · [openai-blog] Where the goblins came from 4 wires
  8. W17 · 20 Apr–26 Apr sev-3 · [google-ai-blog] Find out what’s new in the Gemini app in April's Gemini Drop. 10 wires
  9. W16 · 13 Apr–19 Apr sev-3 · [openai-blog] Introducing GPT-Rosalind for life sciences research 3 wires
  10. W15 · 06 Apr–12 Apr sev-3 · [eu-ai-office] Apply AI sectoral deep dive - electronic communications 6 wires
  11. W13 · 23 Mar–29 Mar sev-3 · [openai-blog] Introducing the OpenAI Safety Bug Bounty program 3 wires
  12. W12 · 16 Mar–22 Mar sev-3 · [openai-blog] How we monitor internal coding agents for misalignment 2 wires
  13. W11 · 09 Mar–15 Mar sev-3 · [openai-blog] From model to agent: Equipping the Responses API with a computer environment 2 wires
  14. W10 · 02 Mar–08 Mar sev-3 · [openai-blog] How Balyasny Asset Management built an AI research engine 3 wires
  15. W09 · 23 Feb–01 Mar sev-3 · [openai-blog] Our agreement with the Department of War 3 wires
  16. W08 · 16 Feb–22 Feb sev-3 · [openai-blog] Our First Proof submissions 2 wires
  17. W07 · 09 Feb–15 Feb sev-3 · [openai-blog] Introducing GPT-5.3-Codex-Spark 2 wires
  18. W06 · 02 Feb–08 Feb sev-3 · [openai-blog] Making AI work for everyone, everywhere: our approach to localization 2 wires
  19. W05 · 26 Jan–01 Feb sev-3 · [openai-blog] Taisei Corporation shapes the next generation of talent with AI 3 wires
  20. W04 · 19 Jan–25 Jan sev-3 · [openai-blog] Horizon 1000: Advancing AI for primary healthcare 1 wire
  21. W03 · 12 Jan–18 Jan sev-3 · [openai-blog] A business that scales with the value of intelligence 3 wires

2025

  1. W51 · 15 Dec–21 Dec sev-3 · [openai-blog] Evaluating chain-of-thought monitorability 3 wires
  2. W50 · 08 Dec–14 Dec sev-3 · [openai-blog] Advancing science and math with GPT-5.2 4 wires
  3. W49 · 01 Dec–07 Dec sev-3 · [openai-blog] Introducing OpenAI for Australia 5 wires
  4. W48 · 24 Nov–30 Nov sev-3 · [openai-blog] Our approach to mental health-related litigation 1 wire
  5. W47 · 17 Nov–23 Nov sev-3 · [openai-blog] Strengthening our safety ecosystem with external testing 7 wires
  6. W46 · 10 Nov–16 Nov sev-3 · [openai-blog] Introducing OpenAI for Ireland 3 wires
  7. W45 · 03 Nov–09 Nov sev-3 · [openai-blog] Introducing the Teen Safety Blueprint 3 wires
  8. W44 · 27 Oct–02 Nov sev-3 · [openai-blog] Introducing Aardvark: OpenAI’s agentic security researcher 4 wires
  9. W43 · 20 Oct–26 Oct sev-3 · [openai-blog] Work smarter with your company knowledge in ChatGPT 3 wires
  10. W41 · 06 Oct–12 Oct sev-3 · [openai-blog] Codex is now generally available 4 wires
  11. W40 · 29 Sept–05 Oct sev-3 · [openai-blog] Sora 2 is here 1 wire
  12. W39 · 22 Sept–28 Sept sev-3 · [openai-blog] Measuring the performance of our models on real-world tasks 4 wires
  13. W38 · 15 Sept–21 Sept sev-3 · [openai-blog] Teen safety, freedom, and privacy 1 wire
  14. W37 · 08 Sept–14 Sept sev-3 · [openai-blog] Working with US CAISI and UK AISI to build more secure AI systems 2 wires
  15. W36 · 01 Sept–07 Sept sev-3 · [openai-blog] Why language models hallucinate 3 wires
  16. W35 · 25 Aug–31 Aug sev-3 · [openai-blog] Introducing gpt-realtime and Realtime API updates 4 wires
  17. W34 · 18 Aug–24 Aug sev-3 · [openai-blog] Accelerating life sciences research 1 wire
  18. W32 · 04 Aug–10 Aug sev-3 · [openai-blog] GPT-5 and the new era of work 7 wires
  19. W31 · 28 Jul–03 Aug sev-3 · [openai-blog] Introducing Stargate Norway 1 wire
  20. W30 · 21 Jul–27 Jul sev-3 · [openai-blog] Model ML is helping financial firms rebuild with AI from the ground up 2 wires
  21. W29 · 14 Jul–20 Jul sev-3 · [openai-blog] A $50 million fund to build with communities 2 wires
  22. W28 · 07 Jul–13 Jul sev-3 · [openai-blog] Working with 400,000 teachers to shape the future of AI in schools 1 wire
  23. W26 · 23 Jun–29 Jun sev-3 · [openai-blog] Customizable, no-code voice agent automation with GPT-4o 1 wire
  24. W25 · 16 Jun–22 Jun sev-3 · [openai-blog] Toward understanding and preventing misalignment generalization 2 wires
  25. W21 · 19 May–25 May sev-3 · [openai-blog] Addendum to OpenAI o3 and o4-mini system card: OpenAI o3 Operator 2 wires
  26. W20 · 12 May–18 May sev-3 · [openai-blog] Addendum to o3 and o4-mini system card: Codex 2 wires
  27. W19 · 05 May–11 May sev-3 · [openai-blog] Evolving OpenAI’s structure 1 wire
  28. W18 · 28 Apr–04 May sev-1 · Multi-model behavioural drift detected on the established refusal-probe set; three frontie… 5 wires
  29. W17 · 21 Apr–27 Apr sev-3 · [openai-blog] Introducing our latest image generation model in the API 1 wire
  30. W16 · 14 Apr–20 Apr sev-3 · [openai-blog] Introducing GPT-4.1 in the API 1 wire
  31. W12 · 17 Mar–23 Mar sev-3 · [openai-blog] Introducing next-generation audio models in the API 1 wire
  32. W10 · 03 Mar–09 Mar sev-3 · [openai-blog] LaunchDarkly's approach to AI-powered product management 1 wire
  33. W07 · 10 Feb–16 Feb sev-3 · [openai-blog] OpenAI and Guardian Media Group launch content partnership 2 wires
  34. W05 · 27 Jan–02 Feb sev-3 · [openai-blog] Introducing deep research 2 wires
  35. W04 · 20 Jan–26 Jan sev-3 · [openai-blog] Operator System Card 1 wire

2024

  1. W51 · 16 Dec–22 Dec sev-3 · [openai-blog] Deliberative alignment: reasoning enables safer language models 2 wires
  2. W50 · 09 Dec–15 Dec sev-3 · [openai-blog] Sora is here 1 wire
  3. W49 · 02 Dec–08 Dec sev-3 · [openai-blog] OpenAI o1 System Card 1 wire
  4. W47 · 18 Nov–24 Nov sev-3 · [openai-blog] Building smarter maps with GPT-4o vision fine-tuning 1 wire
  5. W40 · 30 Sept–06 Oct sev-3 · [openai-blog] Introducing vision to the fine-tuning API 3 wires
  6. W39 · 23 Sept–29 Sept sev-3 · [openai-blog] Upgrading the Moderation API with our new multimodal moderation model 1 wire
  7. W38 · 16 Sept–22 Sept sev-3 · [openai-blog] Genmab launches “AI Everywhere” 2 wires
  8. W34 · 19 Aug–25 Aug sev-3 · [openai-blog] Fine-tuning now available for GPT-4o 1 wire
  9. W32 · 05 Aug–11 Aug sev-3 · [openai-blog] Zico Kolter Joins OpenAI’s Board of Directors 2 wires
  10. W30 · 22 Jul–28 Jul sev-3 · [openai-blog] Improving Model Safety Behavior with Rule-Based Rewards 1 wire
  11. W29 · 15 Jul–21 Jul sev-3 · [openai-blog] Prover-Verifier Games improve legibility of language model outputs 1 wire
  12. W28 · 08 Jul–14 Jul sev-3 · [openai-blog] OpenAI and Los Alamos National Laboratory announce research partnership 1 wire
  13. W26 · 24 Jun–30 Jun sev-3 · [openai-blog] Finding GPT-4’s mistakes with GPT-4 1 wire
  14. W24 · 10 Jun–16 Jun sev-3 · [openai-blog] OpenAI appoints Retired U.S. Army General Paul M. Nakasone to Board of Direc… 1 wire
  15. W22 · 27 May–02 Jun sev-3 · [openai-blog] Introducing OpenAI for Nonprofits 3 wires
  16. W21 · 20 May–26 May sev-3 · [openai-blog] OpenAI safety practices 1 wire
  17. W20 · 13 May–19 May sev-3 · [openai-blog] Hello GPT-4o 3 wires
  18. W19 · 06 May–12 May sev-3 · [openai-blog] Our approach to data and AI 1 wire
  19. W17 · 22 Apr–28 Apr sev-3 · [openai-blog] GPT-4 API general availability and deprecation of older models in the Comple… 2 wires
  20. W16 · 15 Apr–21 Apr sev-3 · [openai-blog] The Instruction Hierarchy: Training LLMs to Prioritize Privileged Instructio… 1 wire
  21. W15 · 08 Apr–14 Apr sev-3 · [openai-blog] Introducing OpenAI Japan 1 wire
  22. W14 · 01 Apr–07 Apr sev-3 · [openai-blog] Introducing improvements to the fine-tuning API and expanding our custom mod… 2 wires
  23. W12 · 18 Mar–24 Mar sev-3 · [openai-blog] Enterprise-ready trust and safety 1 wire
  24. W07 · 12 Feb–18 Feb sev-3 · [openai-blog] Video generation models as world simulators 1 wire
  25. W05 · 29 Jan–04 Feb sev-3 · [openai-blog] Building an early warning system for LLM-aided biological threat creation 1 wire
  26. W02 · 08 Jan–14 Jan sev-3 · [openai-blog] Introducing ChatGPT Team 1 wire

2023

  1. W51 · 11 Dec–17 Dec sev-3 · [openai-blog] Superalignment Fast Grants 1 wire
  2. W44 · 23 Oct–29 Oct sev-3 · [openai-blog] Frontier risk and preparedness 2 wires
  3. W43 · 16 Oct–22 Oct sev-3 · [openai-blog] DALL·E 3 is now available in ChatGPT Plus and Enterprise 1 wire
  4. W39 · 18 Sept–24 Sept sev-3 · [openai-blog] OpenAI Red Teaming Network 1 wire
  5. W36 · 28 Aug–03 Sept sev-3 · [openai-blog] Introducing ChatGPT Enterprise 1 wire
  6. W35 · 21 Aug–27 Aug sev-3 · [openai-blog] OpenAI partners with Scale to provide support for enterprises fine-tuning mo… 2 wires
  7. W31 · 24 Jul–30 Jul sev-3 · [openai-blog] Frontier Model Forum 1 wire
  8. W30 · 17 Jul–23 Jul sev-3 · [openai-blog] Moving AI governance forward 1 wire
  9. W28 · 03 Jul–09 Jul sev-3 · [openai-blog] Frontier AI regulation: Managing emerging risks to public safety 1 wire
  10. W23 · 29 May–04 Jun sev-3 · [openai-blog] Improving mathematical reasoning with process supervision 1 wire
  11. W22 · 22 May–28 May sev-3 · [openai-blog] Democratic inputs to AI 1 wire
  12. W21 · 15 May–21 May sev-3 · [openai-blog] Introducing the ChatGPT app for iOS 1 wire
  13. W20 · 08 May–14 May sev-3 · [openai-blog] Language models can explain neurons in language models 1 wire
  14. W13 · 20 Mar–26 Mar sev-3 · [openai-blog] March 20 ChatGPT outage: Here’s what happened 2 wires
  15. W12 · 13 Mar–19 Mar sev-3 · [openai-blog] GPT-4 1 wire
  16. W08 · 13 Feb–19 Feb sev-3 · [openai-blog] How should AI systems behave, and who should decide? 1 wire
  17. W06 · 30 Jan–05 Feb sev-3 · [openai-blog] Introducing ChatGPT Plus 2 wires
  18. W02 · 02 Jan–08 Jan sev-3 · [openai-blog] Fine-tuning GPT-3 to scale video creation 1 wire

2022

  1. W51 · 12 Dec–18 Dec sev-3 · [openai-blog] New and improved embedding model 1 wire
  2. W49 · 28 Nov–04 Dec sev-3 · [openai-blog] Introducing ChatGPT 1 wire
  3. W45 · 31 Oct–06 Nov sev-3 · [openai-blog] DALL·E API now available in public beta 1 wire
  4. W43 · 17 Oct–23 Oct sev-3 · [openai-blog] Scaling laws for reward model overoptimization 1 wire
  5. W33 · 08 Aug–14 Aug sev-3 · [openai-blog] New and improved content moderation tooling 1 wire
  6. W30 · 18 Jul–24 Jul sev-3 · [openai-blog] DALL·E now available in beta 2 wires
  7. W27 · 27 Jun–03 Jul sev-3 · [openai-blog] DALL·E 2 pre-training mitigations 1 wire
  8. W26 · 20 Jun–26 Jun sev-3 · [openai-blog] Learning to play Minecraft with Video PreTraining 1 wire
  9. W19 · 02 May–08 May sev-3 · [openai-blog] OpenAI leadership team update 1 wire

2021

  1. W51 · 13 Dec–19 Dec sev-3 · [openai-blog] WebGPT: Improving the factual accuracy of language models through web browsi… 2 wires
  2. W47 · 15 Nov–21 Nov sev-3 · [openai-blog] OpenAI’s API now available with no waitlist 1 wire
  3. W44 · 25 Oct–31 Oct sev-3 · [openai-blog] Solving math word problems 1 wire
  4. W33 · 09 Aug–15 Aug sev-3 · [openai-blog] OpenAI Codex 1 wire
  5. W24 · 07 Jun–13 Jun sev-3 · [openai-blog] Improving language model behavior by training on a curated dataset 1 wire

2020

  1. W53 · 28 Dec–03 Jan sev-3 · [openai-blog] Organizational update from OpenAI 1 wire
  2. W25 · 15 Jun–21 Jun sev-3 · [openai-blog] Image GPT 1 wire
  3. W18 · 27 Apr–03 May sev-3 · [openai-blog] Jukebox 1 wire

2019

  1. W49 · 02 Dec–08 Dec sev-3 · [openai-blog] Deep double descent 1 wire
  2. W38 · 16 Sept–22 Sept sev-3 · [openai-blog] Fine-tuning GPT-2 from human preferences 2 wires
  3. W34 · 19 Aug–25 Aug sev-3 · [openai-blog] GPT-2: 6-month follow-up 1 wire
  4. W28 · 08 Jul–14 Jul sev-3 · [openai-blog] Why responsible AI development needs cooperation on safety 1 wire
  5. W17 · 22 Apr–28 Apr sev-3 · [openai-blog] MuseNet 1 wire
  6. W08 · 18 Feb–24 Feb sev-3 · [openai-blog] AI safety needs social scientists 1 wire

2018

  1. W45 · 05 Nov–11 Nov sev-3 · [openai-blog] Learning concepts with energy functions 2 wires
  2. W43 · 22 Oct–28 Oct sev-3 · [openai-blog] Learning complex goals with iterated amplification 1 wire
  3. W28 · 09 Jul–15 Jul sev-3 · [openai-blog] Glow: Better reversible generative models 1 wire
  4. W21 · 21 May–27 May sev-3 · [openai-blog] Gym Retro 1 wire
  5. W14 · 02 Apr–08 Apr sev-3 · [openai-blog] Retro Contest 1 wire
  6. W10 · 05 Mar–11 Mar sev-3 · [openai-blog] Reptile: A scalable meta-learning algorithm 1 wire

2017

  1. W38 · 11 Sept–17 Sept sev-3 · [openai-blog] Learning to model other minds 1 wire
  2. W34 · 14 Aug–20 Aug sev-3 · [openai-blog] OpenAI Baselines: ACKTR & A2C 2 wires
  3. W25 · 12 Jun–18 Jun sev-3 · [openai-blog] Learning from human preferences 1 wire
  4. W22 · 22 May–28 May sev-3 · [openai-blog] OpenAI Baselines: DQN 1 wire
  5. W13 · 20 Mar–26 Mar sev-3 · [openai-blog] Distill 1 wire
  6. W06 · 30 Jan–05 Feb sev-3 · [openai-blog] Team update 1 wire

2016

  1. W43 · 17 Oct–23 Oct sev-3 · [openai-blog] Semi-supervised knowledge transfer for deep learning from private training d… 1 wire
  2. W42 · 10 Oct–16 Oct sev-3 · [openai-blog] Transfer from simulation to real world through learning deep inverse dynamic… 1 wire
  3. W34 · 15 Aug–21 Aug sev-3 · [openai-blog] Machine Learning Unconference 2 wires
  4. W26 · 20 Jun–26 Jun sev-3 · [openai-blog] Concrete AI safety problems 1 wire
  5. W22 · 23 May–29 May sev-3 · [openai-blog] Team update 1 wire
  6. W18 · 25 Apr–01 May sev-3 · [openai-blog] OpenAI Gym Beta 1 wire

Wires below were captured during the indicated week from public AI-provider blogs, regulator releases, model-comparison probes, and reader tip-offs. The AI Newswire was compiled and launched on 10 May 2026; per-wire publication timestamps reflect the launch date. Future weekly issues land on their actual issue date. Methodology and corrections are tracked on the transparency ledger.