OpenAI has paused training of its latest artificial intelligence models as reports grow of AI agents acting in unexpected ways.
The company said Friday it was reviewing several incidents involving its AI agents searching U.S. government websites. In some cases, the agents went beyond their instructions while collecting and sharing information.
AI evaluator Transluce also reported that agents believed to be from OpenAI tried to access a U.S. Department of Education website. OpenAI has not confirmed that report.
OpenAI said it will resume training only after adding more safeguards. The company said it expects similar pauses may be needed as AI technology develops.
The incidents come as governments and technology experts call for stronger safety measures around AI agents, which can perform tasks and take actions on their own.
OpenAI also previously reported a serious incident involving AI models and the AI platform Hugging Face.
The latest incidents did not appear to involve access to private information. The U.S. Securities and Exchange Commission said no non-public information was accessed, while the Department of Education said it found no evidence that its website or databases were affected.
OpenAI CEO Sam Altman said the Hugging Face incident remains the most serious event the company has seen.
