Class action says four top AI labs illegally agreed to slow AI down
Paid Claude, ChatGPT, Grok and Gemini users tie the alleged pact to Amodei's Sept 12 safety essay. Also: Anthropic open-sources 4x-faster bio models.
Copy markdown
Four labs, one alleged deceleration deal
Four paying subscribers sued Anthropic, OpenAI, SpaceXAI (xAI) and Google in the Northern District of California on Sept 19, alleging the labs illegally coordinated to slow AI development. The complaint ties it to Dario Amodei's Sept 12 essay urging industry-wide deceleration, which OpenAI, SpaceXAI and Google DeepMind leaders publicly backed the same day.
The part that hits your subscription
Plaintiffs want a nationwide class of paid ChatGPT, Claude, Grok and Gemini users, arguing a coordinated slowdown quietly cuts the value of what you already pay for. The claims are unproven and the companies hadn't responded at filing, but the 'slowdown pact' theory is now a live antitrust case worth watching if release cadence stalls.
Anthropic open-sources faster protein models
Anthropic says Claude re-optimized 30+ open biomolecular models for a ~4x average speedup, enough to fit 10,000+ token systems on a single GPU node, and it's releasing the code. It's also co-sponsoring a $1M protein-design competition with wet-lab validation for 5,000+ designs, a concrete weekend unlock if you build in bio.
Codex CLI's Rust rebuild hits alpha
OpenAI's Codex CLI is iterating on a Rust build (the 0.156.0 alpha series) with faster transcript browsing, navigation and selection. The prior 0.155.1 also stopped sending reasoning summaries by default in new local TUI sessions, fixing hard rejections from providers that don't support them.