OpenAI slows some AI development to strengthen safeguards
OpenAI has reduced the pace of parts of its artificial intelligence development while it works to improve security and safeguards. The company announced the change on Tuesday, as it faces a possible initial public offering and strong competition from Anthropic, Chinese developers and rivals releasing open-weight models.
The company said it was pausing reinforcement-learning training for two weeks on its latest models intended for deployment. It also said that its largest planned frontier reinforcement-learning run would remain delayed, without specifying a new schedule.
The move puts into practice a position long promoted by AI safety advocates: that companies should be prepared to slow or temporarily step away from the race to develop increasingly capable systems. OpenAI’s decision comes despite the commercial and competitive pressure to accelerate development. The announcement did not provide further details about the security work or explain when the delayed training would resume.
Written by Sakhanda Wire AI from reporting by The Verge. Facts, quotes and full context are in the original article below.