Anthropic says in its amended IPO filing that it has met its safety threshold for AI-accelerated research
"We made a determination that we have met this threshold," the prospectus says, basing the call on expected future models and its rate of progress rather than on current models; crossing the line means tougher safeguards, not a shutdown.
Why it matters
It is the first time a frontier lab has formally declared that AI-driven acceleration of its own research has crossed one of its safety thresholds. Under Anthropic's Responsible Scaling Policy, the line is reached when AI does enough of the company's research to roughly double how fast its models improve, packing two years of 2018–2024 progress into one. The call rests on expected future models rather than measured results. METR separately estimated that Anthropic's models have improved nearly twice as fast since October 2025. Watch for the further reports Anthropic says it will publish and any policy changes to keep humans in control.
Key facts
- Disclosed in Anthropic's amended IPO prospectus on 10 November 2026
- "We made a determination that we have met this threshold"
- Threshold: AI research roughly doubling how fast models improve
- Based on expected future models, not observed current models
- METR: models improving nearly twice as fast since October 2025
- Same filing set the price range at $70 to $78 a share