Today's AI page · AI digest archive

AI digest · 2026-10-05

The 10 AI stories that the most sources reported on this day, each with the reports it was drawn from.

  1. Trump rebrands AI as "super intelligence" and launches safety force — Major tech CEOs signed a White House pledge while a new government unit coordinates AI policy

    Sources:[12] techcrunch.com[13] techcrunch.com[15] the-decoder.com

  2. OpenAI safety researcher quits and publicly sounds the alarm — David Robinson resigned and wrote in The Atlantic about concerns with the company's safety reporting practices

    Sources:[5] theverge.com

  3. Google introduces RRSI to stop self-improving AI agents from cheating on tests — The method boosts unseen benchmark scores by up to 4.7 points while cutting token use by ~30%

    Sources:[16] the-decoder.com

  4. Google locks free Gemini users to weakest model, charges for better tiers — Starting October 2026, only Flash-Lite remains free while Flash and Pro become paid-only

    Sources:[19] the-decoder.com

  5. Apple tightens macOS disk access controls due to risks from AI agents — Broader file and data access by increasingly capable agents prompted new permission safeguards

    Sources:[11] techcrunch.com

  6. Meta open-sources code for building custom Muse AI gadgets — Developers can now create their own hardware featuring Meta's new AI agent on displays, HDMI sticks, and more

    Sources:[6] theverge.com

  7. NASA and IBM release open-source Lunar Foundation Model — Trained on 17 years of orbiter data, it improves prediction of polar ice deposits for lunar science

    Sources:[17] the-decoder.com

  8. Sean Parker rebuilds Stability AI around music with industry backing — The Napster founder returns to music tech, this time with labels' investment and approval

    Sources:[9] techcrunch.com

  9. Capcom prepares for game development "together with AI" — The studio detailed plans to evolve its RE Engine for AI-assisted creation at its developer conference

    Sources:[3] theverge.com

  10. Chinese AI models heavily parrot state doctrine on sensitive topics — An Aleph Alpha study found only 17-41% of answers on political questions were balanced

    Sources:[18] the-decoder.com