Major AI labs are negotiating deployment timelines directly with regulators

Anthropic's deferral of Claude Mythos, a cybersecurity-focused model, is now part of European Commission review. OpenAI, Google DeepMind, and Anthropic have all adjusted release schedules following government engagement. The shift is structural: labs now negotiate capability access upstream rather than releasing first and facing rules afterward. Whether this represents genuine safety governance or regulatory capture remains contested.