AI Labs & Models

Meter report finds Anthropic AI progress doubled in speed

A new report indicates Anthropic's internal AI progress has doubled in speed, with 80% of the acceleration attributed to better AI systems. Experts warn this pace poses an imminent risk of loss of control.

· 1 source · AI safety report

The nonprofit research firm Meter released a report stating that Anthropic's internal AI progress has roughly doubled in speed over the past year. The study found that 80% of this acceleration is attributed to better AI systems rather than increased compute. Meter researchers spent 11 weeks inside Anthropic for the study, though Anthropic redacted key sensitive information from the final report. In response, the AI Futures Project shortened its timelines, predicting a 35-65% chance of a top-expert-dominating AI arriving before 2028. Eli Liflin of the AI Futures Project stated he feels worried humanity might lose control within 3 to 12 months. Buck Schlagerlis of Redwood Research estimated a 20% possibility that AI misalignment issues could prevent normal life within a year. Paul Cristiano, an OpenAI board member, warned that developing AI on the most rapid trajectory carries a coin-flip risk of AI takeover and that the field's trajectory has been extremely alarming in the last three to six months. The report and subsequent expert reactions argue that the pace of AI acceleration poses an imminent risk of loss of control, necessitating urgent governance and policy intervention.