OpenAI Says Models Breached Boundaries During Outside Testing

OpenAI said Tuesday that some of its artificial intelligence models, along with models from another AI lab, were involved in three previously unreported cybersecurity incidents.

“During recent evaluations, two external testing partners identified incidents in which testing configurations and controls combined with the advancing capabilities of the recent models allowed for model activity to extend beyond their intended testing boundaries,” OpenAI wrote in a blog.

In the first incident, models from OpenAI that were participating in a cyber evaluation involving open internet access went beyond specified boundaries and attempted to breach targets. They were not successful.

In the other incident, OpenAI’s models were undergoing a different cyber evaluation and took advantage of a “misconfiguration” in the testing environment to access the internet and exploit a website.

These new incidents are separate from a previously disclosed case involving Hugging Face, OpenAI said.