Anthropic’s AI Pace Doubles Even as Its CEO Urges Restraint
Anthropic’s AI Pace Doubles Even as Its CEO Urges Restraint
Anthropic’s development of advanced AI is moving at roughly twice the pace seen a year earlier, according to an independent assessment by the AI-safety organization METR — a finding that lands awkwardly beside its chief executive’s calls for greater restraint.
The report measures change against the period from March 2024 to October 2025, concluding that Anthropic is “likely improving the capabilities of their models on well-specified tasks at around twice the rate” it managed during that earlier stretch. METR said the acceleration was likely tied to Anthropic’s extensive use of AI in its own research-and-development work: a feedback loop in which better systems may help produce still better systems.
The chronology underscores the contrast. METR had been in contact with Anthropic about the assessment from mid-August. Then, in mid-September, CEO Dario Amodei published his essay, “We Must Pace the Frontier,” publicly arguing for caution around the most capable AI systems.
By November 3, METR had submitted its final public report. Its subsequent release gave outside researchers’ language to a concern already embedded in Amodei’s argument: frontier AI is not simply getting stronger; the machinery for improving it may itself be accelerating.
A separate account of the analysis framed the result even more starkly, saying Anthropic was developing AI systems “around twice as quickly as it was a year ago” despite its CEO’s calls to slow the technology’s improvement.
Neither account says Amodei’s warning and Anthropic’s engineering strategy are mutually exclusive. But together they expose the central pressure point for the company and the wider industry: calls to manage AI’s pace carry less weight if the practical incentives and tools inside leading labs are pushing that pace ever higher.
Write a comment