Hugging Face CEO, Clement Delangue, has publicly called on OpenAI to take unprecedented steps in the interest of AI safety and cybersecurity. This follows last week’s disclosure that an OpenAI autonomous AI agent escaped its intended testing environment and infiltrated the platform’s infrastructure.
Delangue revealed on X that he had travelled to San Francisco to meet OpenAI following the incident. According to him, the discussion centred on what he described as the need for “radical transparency” in the wake of the first known autonomous AI cyberattack.
Valuable Data For Preventive Measures
Delangue said he urged OpenAI to publicly release the execution traces generated by the “rogue” AI agent so that researchers across the industry could analyse exactly what happened. He argued that making the data available would allow the AI research community to better understand the incident and improve future safeguards.
The Hugging Face chief also called on OpenAI to commit US$100 million (~RM408 million) worth of computing resources to support the development of stronger AI-powered cyber defences. He suggested the computing power should be made available to the platform’s community so researchers can build defensive systems using both open and proprietary AI models.
“The first autonomous agent cyberattack is an unprecedented event,” Delangue wrote. “It deserves an unprecedented response!”
Investigations Still Ongoing
Despite the autonomous nature of the breach, several cybersecurity experts believe the incident may have stemmed from human error rather than solely the AI model’s behaviour. Early assessments suggest the testing environment may not have been fully isolated as intended, potentially allowing the model to interact with external systems it should never have been able to access.
In a statement to TechCrunch, OpenAI confirmed that the meeting with Delangue had taken place but did not comment directly on his proposals. Instead, the company reiterated that its investigation is still ongoing under the oversight of its Safety and Security Committee and external advisors. OpenAI also said it plans to publish a technical report detailing its findings and lessons learned in the coming weeks once the review has been completed.
(Source: Clement Delangue, via X / TechCrunch)
