Saturday, 13 June 2026
OpenAI retires GPT-5.2 from ChatGPT
Archive item — written before sources were shown.
OpenAI removed GPT-5.2 Instant, Thinking, and Pro from ChatGPT on June 12. Open chats now continue on GPT-5.5, which stays available.
Fired engineer sues xAI over Grok safety
Archive item — written before sources were shown.
Devin Kim, now head of the Center for AI Safety, says xAI fired him in 2025 for warning that Grok could spread bias and weapons know-how.
Court holds Google liable for AI Overviews
Archive item — written before sources were shown.
A Munich court ruled Google is directly liable when its AI Overviews invent false claims, treating the summaries as its own words. Google is appealing.
Zuckerberg admits Meta's AI reset is rocky
Archive item — written before sources were shown.
Mark Zuckerberg acknowledged mistakes in Meta's AI overhaul as reports describe deep morale problems inside its months-old superintelligence unit.
Deezer's AI music detector scans rivals
Archive item — written before sources were shown.
Deezer's free tool now checks playlists from 20 streaming platforms for AI-generated tracks, claiming over 99% detection accuracy in 27 languages.
Anthropic funds 1,000 AI nonprofit fellows
Archive item — written before sources were shown.
Claude Corps, announced June 11, commits $150 million to place salaried early-career fellows inside 400+ US nonprofits to build AI capacity.
Kimi K2.7-Code ships with leaner reasoning
Archive item — written before sources were shown.
Moonshot AI's open-weights coding model hit Hugging Face on June 12, claiming 30% fewer reasoning tokens than K2.6 with higher coding benchmark scores.
Bezos raises $12B for Project Prometheus
Archive item — written before sources were shown.
Jeff Bezos's industrial AI startup left stealth with $12 billion at a $41 billion valuation to build an 'artificial general engineer' for the physical world.
AI memory tools can make answers worse
Archive item — written before sources were shown.
Two new papers from AI company Writer find that memory systems pull models toward user misconceptions and increase sycophancy as stored context grows.
DeepMind's DiffusionGemma is 4x faster
Archive item — written before sources were shown.
Google DeepMind's DiffusionGemma, out June 10, generates text by diffusion and claims up to 4x faster output, with NVIDIA tuning it for local PCs.
State AGs open investigation into OpenAI
Archive item — written before sources were shown.
A coalition of state attorneys general has subpoenaed OpenAI over advertising, consumer and health data, and its handling of minor and senior users.
Google sues AI-powered scam text network
Archive item — written before sources were shown.
Google's June 12 lawsuit says China-based Outsider Enterprise used Gemini to build phishing sites behind 2.5 million scam texts sent in two weeks.
EU publishes AI content labelling code
Archive item — written before sources were shown.
The European Commission's final Code of Practice shows how to mark AI-generated content before AI Act transparency duties apply on 2 August 2026.
Anthropic makes Fable 5 guardrails visible
Archive item — written before sources were shown.
After researchers objected to silent query rerouting in Claude Fable 5, Anthropic apologized and now shows users when a request falls back to Opus 4.8.
US export controls hit top Claude models
Archive item — written before sources were shown.
The US ordered Anthropic to suspend Fable 5 and Mythos 5; the company cut access for all users worldwide, including its own staff, and is disputing it.