Anthropic is rapidly accelerating AI development amid slowdown calls, outside research group says

Anthropic is developing its AI systems about twice as fast as it was a year ago, according to an analysis released by an independent monitoring group on Thursday. The findings come as the San Francisco company prepares for a major stock offering and faces heightened debate about whether advanced models could evade human control with catastrophic risks. Researchers from Model Evaluation and Threat Research, based in Berkeley, reviewed internal data and interviewed staff over roughly 11 weeks, from mid-August to late October. They assessed whether capability growth accelerated and how much could be attributed to AI. The group said it had unprecedented access, though Anthropic’s redactions and internal uncertainty limited some estimates. Anthropic did not immediately comment.