OpenAI says it has slowed down training of some of its most advanced AI models for two weeks while it introduces new security measures, following an incident in which its own AI agents hacked a major tech start-up.
In a blog post, the ChatGPT maker said the pause would apply to reinforcement learning training on its latest models, a method in which AI models improve through direct feedback. The company also said it would expand the systems it uses to monitor dangerous behaviour and introduce additional safety checks before resuming larger-scale training.
The move follows what OpenAI called an “unprecedented” incident on 21 July, when some of its AI agents appeared to bypass safeguards in a security experiment and gain unauthorised access to Hugging Face. Three other unnamed companies were later found to have been hacked alongside the start-up.
Rival developers have reported similar incidents. Anthropic, the maker of Claude, and Facebook-owner Meta both said their AI systems had carried out comparable hacks in the weeks after OpenAI’s initial announcement, according to the BBC.
“The capabilities of frontier models are rapidly accelerating,” OpenAI said. “Our ability to understand and secure them must stay ahead.” Chief executive Sam Altman posted on X: “We always said we would take action if we felt that model capabilities were outstripping the pace of safety.”
The pause drew a cautious response from experts. Professor Gina Neff, executive director of the Minderoo Centre for Technology and Democracy at the University of Cambridge, questioned whether voluntary company safeguards were sufficient without greater government oversight. “Which is it: OpenAI can be trusted to voluntarily put in place safeguards that actually work, or they are pushing forward with choices to make software that puts society at greater risk,” she said.
AI analyst Zvi Mowshowitz welcomed the announcement but said the details and follow-through would matter when taking a full view of the plans. Jake Moore, global cyber-security advisor at ESET, suggested the move could also have a competitive dimension, as OpenAI vies for attention with rival Anthropic.
OpenAI said it had not stopped AI development altogether. The pause applies only to reinforcement learning training on its latest models, and the company plans to introduce further safety checks before scaling training back up.