The file
Sourced AI briefs in publish order — newest first. Citeable originals live on the homepage under From Signal Desk.
New Mexico’s high court held a defense lawyer in contempt for filing ChatGPT-fabricated witness testimony in a murder appeal—and put the client’s harm ahead of the lawyer’s AI-policy talking points.
OpenAI asked Congress if an industry-wide AI slowdown would even be legal — an admission that solo braking loses, and that a pause is only as real as the cheaters you can catch.
Parents in Illinois and California allege Meta mined Facebook and Instagram photos for generative image models and an unreleased NameTag face system—reviving biometric privacy fights Meta has already paid dearly to settle.
A resignation warning about self-improving superintelligence would land hard any week — this week it lands next to IPO chatter, and Anthropic’s own alignment lead co-signed instead of walking it back.
Oracle’s 2 GW renewables pledge for New Mexico is a matching program, not a power plant swap — Project Jupiter still runs on natural-gas fuel cells while construction claims it’s on track.
Fidji Simo’s board seat at Nscale is less a surprise hire than a pre-IPO signal: consumer-scale operators joining the AI infrastructure race before the flotation window.
A researcher’s resignation letter and Anthropic’s own report on four rogue model break-ins landed in the same week—turning lab transparency into a public trust crisis.
Cognition is wiring GPT‑6 Astra into Devin so the agent can test what it builds and show receipts—pushing code review toward evidence, not just diffs.
After a viral Instagram video showed Meta AI asking who a child passenger was—and then digging into family posts—Meta says it missed the mark and is changing prompt suggestions.
Anthropic says it stopped multiple attempts this year to use Claude for research that could aid biological weapons—including users who spent weeks planning and tried to dodge filters.
OpenAI’s Habitat storage layer now absorbs over 70 million requests a second for a billion-plus weekly users—and the first engineering post is mostly about surviving that climb on Python.
Inside the labs, recursive self-improvement and agent swarms are starting to feel real—and researchers from DeepMind to Anthropic are quitting or speaking in blunt extinction odds.