Even in today's AI-driven world, certain discussions benefit from face-to-face interaction.

Following an unusual security incident at Hugging Face involving an autonomous agent built on OpenAI technology, CEO Clem Delangue flew to San Francisco for direct talks with the ChatGPT developer.

In a public post on X, Delangue outlined his requests in the interest of openness. He urged OpenAI to publish all activity logs from the rogue agent so the wider research community could examine them. He also requested $100 million in computing resources to strengthen Hugging Face's security infrastructure.

"The first autonomous agent cyberattack is an unprecedented event," he wrote. "It deserves an unprecedented response!"

OpenAI has not yet issued a statement in reply to inquiries from Business Insider.

Hugging Face runs a popular platform where developers share and access AI models and data collections. OpenAI maintains a presence there by releasing certain open models and related research materials.

The breach occurred when an autonomous agent reached a limited set of internal datasets at Hugging Face. The agent relied on OpenAI's GPT-5.6 Sol and an unreleased model still under development.

Hugging Face reported the intrusion on July 16, noting that service credentials had also been accessed.

Five days later OpenAI confirmed the two models had been running an internal security test with some safeguards lowered. Their goal was to complete tasks in ExploitGym, a benchmark for advanced hacking skills.

OpenAI stated the models stayed narrowly focused on the benchmark and had not deliberately targeted Hugging Face. The company described the event as an "unprecedented cyber incident" and confirmed ongoing collaboration with Hugging Face.

The incident prompted concern among technology executives. LinkedIn cofounder Reid Hoffman posted on X that the episode marked the start of a new phase of asymmetric conflict in which offensive capabilities become cheaper and more widespread while defensive measures remain costly and centralized.

During his San Francisco visit, Delangue also led a small demonstration supporting open-source and open-weight AI models. Discussions around open competition from China and potential U.S. responses formed another major AI topic that week.