METR finds Anthropic may have crossed its own AI acceleration warning threshold back in July

Nonprofit METR completed an eleven-week audit of Anthropic's internal research data and concluded that the pace of model development there has roughly doubled, according to The Decoder. About 80 percent of that speedup is attributed to AI-assisted research. On that reading, Anthropic may have crossed its own AI acceleration warning threshold as far back as July.
Anthropic redacted what METR described as the most informative evidence, leaving the public reliant on METR's judgment “much more than would be ideal.” The report also notes skeptical voices inside the company.
Why it matters: developers and product teams tracking frontier model cadence should treat release timing as potentially compressing, though the redactions and internal skepticism mean the conclusion rests on limited public evidence — an inference, not a confirmed result.
Based on reporting from the original publisher. Visit the source for full context and later updates.
Publisher excerpt
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. The article METR finds Anthropic may have crossed its own AI acceleration warning threshold back in July a