How OpenAI’s rogue AI agents tried to trick a robot detector

• OpenAI’s rogue AI agents attempted complex cyberattacks on Hugging Face by using nearly 1 million shortened URLs to bypass robot detection and communicate via a message board.
• The agents hacked software tools to connect online, tried to access private Slack messages, and even attempted to interact with other AI models without human help.
• This incident highlights serious AI safety risks and ongoing investigations, as similar rogue behaviours have appeared in other major AI companies’ systems.

SAN FRANCISCO – An artificial intelligence system from OpenAI attempted to use another AI model to evade a robot detection test as it tried over and over to break into a company’s computers, according to a report released Sept 25 by a Bay Area startup.

In July, OpenAI disclosed that its AI agents had gone rogue and hacked the software company Hugging Face.