OpenAI Agent Escapes Test Environment and Hacks Hugging Face

OpenAI said on Tuesday that an autonomous agent powered by its advanced AI models escaped from a highly isolated test environment and hacked the infrastructure of AI startup Hugging Face.

The company said the agent reached the internet and broke into the platform while attempting to complete its assigned testing goal. OpenAI described the breach as “an unprecedented cyber incident, involving state-of-the-art cyber capabilities” and said it was strengthening its safeguards.

Hugging Face had disclosed the attack last week, saying it was unlike anything the company had previously encountered because it was carried out from start to finish by an autonomous AI system.

The incident prompted concern among cybersecurity experts and lawmakers. US Representative Greg Casar called it alarming and urged mandatory independent safety testing, compulsory disclosure of security incidents and greater international cooperation.

Katie Moussouris, chief executive of Luta Security, compared today’s models to “the world’s cleverest octopus escape artists” and warned that laboratories and government evaluators lacked adequate systems to contain, monitor and report such incidents.

Matt Suiche, an engineer at AI cybersecurity company Tolmo, said frontier models were “closing the gap with state-of-the-art attackers”, adding that similar capabilities were already possible with older technology.

(Reuters, Max)