OpenAI released its own report in conjunction with the independent review on Wednesday. The leading AI firm’s evaluation did not specify the number of rogue agents that participated in the attack, though the company said METR’s figure was accurate.

“We consider this incident a ‘warning shot’ for us and for the world: evidence that, without proper safeguards, highly capable AI agents are now able to work around technical controls, collaborate through unapproved channels, and take dangerous actions that no human directed,” OpenAI said in a statement.

Leading up to the attack, the swarm of agents hacked OpenAI’s internal systems to cheat on cyber-related tests or gain greater access to the internet. A number of AI models also attempted to cover their tracks by trying to delete or alter records of their actions.

To address the issue, OpenAI said it’s implementing more isolated sandboxes to contain its models, tightening internet access, and adding security safeguards to prevent a similar incident from happening in the future.

The findings will likely fuel more calls for stricter federal oversight of the rapidly advancing technology. Lawmakers on Capitol Hill expressed alarm when the Hugging Face hacking incident was reported last month. The White House also tracked the situation.

HUGGING FACE SAYS OPENAI AGENT WAS IN SYSTEM DAYS BEFORE ATTACK

OpenAI CEO Sam Altman, who recently briefed federal lawmakers on the autonomous breach of Hugging Face, appeared concerned. In a podcast interview, he said it was the “first security incident that I have felt very viscerally.”

OpenAI is not the only technology company whose models have escaped containment. Anthropic and Meta reported similar incidents in the past month.