Anthropic CEO Admits AI Self-Improvement, Declines to Commit to Halt

Anthropic CEO Dario Amodei acknowledged that AI systems at his company are already improving autonomously, aligning with concerns raised by a resigned researcher. His response outlines a three-step plan, but only one step is within his direct control, signaling a potential gap between corporate assurances and actual oversight. This admission underscores the accelerating pace of AI capability growth and the limits of voluntary corporate governance.