Anthropic’s AI development pace doubles, METR analysis finds
The analysis, based on 11 weeks of internal data and employee interviews, suggests Anthropic’s progress stems largely from AI-assisted development, with its Claude models helping build newer versions. METR noted no clear bottlenecks in compute or data, raising concerns about unchecked acceleration and potential loss of human oversight as systems improve recursively.
Key points
- METR estimates Anthropic’s AI model improvements doubled in speed over the past year, based on internal data and interviews
- AI-assisted development—specifically Claude models—likely drives the faster pace, with no major compute or data limits identified
- CEO Dario Amodei urged caution in a September essay, warning recursive self-improvement could outrun human control
The report coincides with Anthropic’s upcoming IPO and CEO Dario Amodei’s call for slower, safer AI advancement. Amodei warned recursive self-improvement could outpace human control, advocating for external evaluators like METR. METR itself has worked with Anthropic, OpenAI, and other labs for years, relying on donated access and tokens for evaluations. The findings highlight growing industry debates over safety and governance as AI labs push forward.
Anthropic AI development speed doubles, analysis finds, as CEO urges slower safety-focused pace
goodreturns.in · 5 November 2026
Loading the full article…
This text was published by goodreturns.in and written by Vasant Shah. It is reproduced here with attribution so you can read it in full; the rights remain with the publisher. Read it at the source ↗
Coverage and discussion
4sources- Anthropic prepares report on AI progress and safety practicesPress · cryptobriefing.com ·
- Anthropic’s new AI model achieves over 50% on all benchmarks in METR analysisPress · cryptobriefing.com ·
- METR says Anthropic’s AI models are improving at roughly twice the historical pacePress · cryptobriefing.com ·
The headline, key points and digest above were generated by Digest AI's editorial model from the linked sources. Automated summaries can contain errors: the sources are the record. Spotted a mistake? Tell us. Published by Martin K., who runs Digest AI and handles corrections.
More in Policy & Regulation
All →- Microsoft patents AI that could edit live speech and match lips · 2 src
- Republicans ask Nvidia CEO to brief Congress on alleged Chinese spy ties · 1 src
- Google says schools must accept AI features to use Search starting next year · 39 src
- SpaceX shares drop 7% amid AI and tech market sell-off after Democratic midterm wins · 8 src
- OpenAI pledges to avoid NDAs and cover electricity costs for data centers after election backlash · 5 src
Comments
via GitHub Discussions