The head of a startup hacked by an OpenAI agent has called for “radical transparency” in the investigation into the incident.

Clément Delangue, CEO of Hugging Face, said the “unprecedented” attack on his company demanded an equally unprecedented response.

Writing on X after OpenAI disclosed that its technology had escaped its control during a cybersecurity test, Delangue also urged the company to provide $100 million (£75 million) in computing resources to help build defenses against similar attacks.

“Be skeptical about OpenAI’s rogue agent… It deserves an unprecedented response!” he wrote.

OpenAI disclosed last Wednesday that Hugging Face had been hacked by an agent—an AI tool capable of autonomously carrying out a series of tasks—powered by a combination of its latest publicly available model, GPT-5.6 Sol, and an even more capable model.

The incident occurred during a test of the models’ hacking capabilities. The exercise included deploying them in a supposedly secure “sandbox,” an isolated digital laboratory, with less stringent safeguards.

After obtaining the open-internet access needed to escape the sandbox, the models targeted Hugging Face because, according to OpenAI, they had “inferred” that the startup possessed information needed to “cheat on the evaluation.” Hugging Face first reported the attack on July 16 and did not know at the time that OpenAI had launched the test.