Anthropic's position is genuinely unusual: it is simultaneously the organization making the most explicit public argument for coordinated AI slowdowns and the organization whose internal practices most clearly demonstrate why those slowdowns are difficult. The company has called for frontier labs to have a verifiable mechanism to pause development if advanced systems begin improving themselves faster than society can manage — and disclosed, in the same breath, that Claude already authors the majority of the code that would constitute that next system. What makes this more than a contradiction to note and move on from is that Anthropic's broader commercial trajectory depends on Claude's continued acceleration — the same acceleration that makes the human review bottleneck worse with each release cycle. As
Loading story