DigiNews

Summary

TIME reports that Anthropic is dropping the core pledge of its flagship Responsible Scaling Policy, which once forbade training models unless safety measures could be guaranteed in advance. The company now commits to greater transparency about safety risks, regular risk reporting, and frontier safety roadmaps, while preserving incentives to pursue responsible research in a fast-moving field. Critics worry the shift could let dangerous capabilities emerge gradually without clear alarm points, even as Anthropic frames it as a pragmatic answer to competitive pressure and evolving political realities.