Research · High impact
METR finds Anthropic may have crossed its own AI acceleration warning threshold back in July
The Decoder · By Maximilian Schreiner ·
The Brief
- Nonprofit METR spent eleven weeks auditing Anthropic's internal research data.
- METR concluded that the pace of Anthropic's model development has roughly doubled.
- About 80 percent of the speedup is attributed to AI-assisted research.
- Anthropic redacted the most informative evidence, according to The Decoder.
Why it matters
If AI-assisted research is driving much of frontier model development, the release cadence of frontier models may keep accelerating, and the redacted evidence limits how well outside observers can verify the claims.
Who’s affected
- Teams relying on Anthropic models
- Researchers tracking AI development pace and AI-assisted research
- #anthropic
- #metr
- #ai-acceleration
- #ai-safety
- #ai-research
- #model-development
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…
Timeline
The Decoder reports METR's findings that Anthropic may have crossed its own AI acceleration warning threshold
The Decoder
METR finds Anthropic may have crossed its own AI acceleration warning threshold back in July