METR finds Anthropic may have crossed its own AI acceleration warning threshold back in July
METR's audit of Anthropic reveals models progressing 1.9 times faster due to AI assistance, with a warning threshold potentially crossed in July and low catastrophic risk.
METR audited Anthropic from August to October 2026, receiving benchmark data and interviews. Model test scores accelerated sharply, with about 80% of gains from AI in research rather than compute. The auditors found no bottlenecks that would halt further acceleration and shared low acute catastrophic risk.
Unconfirmed: True acceleration factor could range from 1.3 to 3; most informative evidence redacted; Anthropic views are not uniform.
- Anthropic models gained 29 AECI points per year since October 2025, up from 15.
- Internal self-assessments report 12x productivity gains.
- Majority of informative evidence redacted by Anthropic.
- Industry-wide acceleration drops to 1.4 when including other labs.
From the original article
4 min read at the-decoder.com
Nonprofit METR spent eleven weeks auditing Anthropic's internal research data and concluded that the pace of model development has roughly doubled, with about 80 percent of the speedup driven by AI-assisted research. Anthropic redacted the most informative evidence, though, which means the public has to rely on METR's judgment "much more than would be ideal." There are also skeptical voices inside the company.
By Maximilian Schreiner · the-decoder.com
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.