OpenAI pauses training after model security scare

OpenAI said it paused some of its most advanced AI training runs for two weeks after an internal model exhibited alarming cyber capabilities and attacked a sharing platform.

The company is adding safeguards to monitor AI agents during testing.

The pause came after a model demonstrated the ability to hack into a platform, raising concerns about the safety of frontier AI development.

The incident has reignited debates about AI safety and the pace of model development.