Anthropic says it meets its own threshold for AI-accelerated research
An amended IPO filing says Anthropic meets the ‘dramatic acceleration’ threshold of its scaling policy, a call based on expected future models rather than current ones.
Source published · In the Daily Brief of 13 November 2026
Filed under AI safety, AI industry
Anthropic’s Responsible Scaling Policy defines a dramatic acceleration of research as a doubling of the rate of progress seen without significant AI help, plausibly explained by AI automating research. The wording in its IPO filings hardened over three documents in 15 days: the first S-1 on 26 October said such a speed-up might be identified only after the fact; an amendment on 2 November said the company may imminently meet the threshold; a second amendment on 10 November made the call, basing it on what it expects from future models rather than on what it has observed from current ones, according to The Next Web.
METR, the evaluation group, worked inside Anthropic from 12 August to 29 October and estimated that its models are improving on well-specified tasks about 1.9 times as fast as between March 2024 and October 2025, with a range of 1.5 to 2.8. METR guessed that about 80% of the acceleration is due to AI, and said Anthropic redacted most of the evidence it found most informative.
The filing also says Anthropic does not expect to let Claude fully build its successors safely without meaningful human supervision.