Sam Altman told OpenAI employees the company is prepared to slow development of next-generation models, but only if competing labs commit to the same pace. Bloomberg reports that OpenAI has already paused some pretraining runs on safety grounds. Altman has previously discussed the idea of a controlled development tempo with officials at the White House. The initiative comes amid a wave of safety-focused departures from OpenAI, Anthropic, and Google DeepMind. This week, OpenAI chief scientist Jakub Pachocki publicly called on the industry to coordinate and accept voluntary release pauses.

Related: OpenAI Slows Model Development Over Cyber Capabilities, OpenAI Shares Safety Lessons from Long-Horizon Model Testing

Bloomberg — OpenAI Is Open to Slowing Cutting-Edge AI, CEO Sam Altman Tells Staff