Bottom line: Clément Delangue is not treating the recent breach involving OpenAI's models as a typical security incident. Rather than limiting his response to internal fixes or legal action, the Hugging Face CEO is publicly calling for two specific concessions: full disclosure of what happened within the systems and a major commitment of computing power to build defenses.
At the center of his response is a call for what he describes as "radical transparency." Delangue wants OpenAI to release complete traces of the models' activity during the incident, including the steps they took and the systems they accessed. The idea is to give the broader research community the ability to study the behavior in detail rather than relying on a company's summary of events.
His second demand is more concrete. Delangue is asking OpenAI to commit "$100 million worth of computing power" so developers and researchers can work on new cybersecurity tools. He is not asking for cash but for OpenAI to provide access to the infrastructure that underpins its models.
"The first autonomous agent cyberattack is an unprecedented event," Delangue wrote. "It deserves an unprecedented response!"
Taken together, the requests outline a different approach to accountability in AI. Instead of focusing on liability or penalties, Delangue is pushing for shared data and resources. The goal, as he frames it, is to treat the incident as a problem for the entire field rather than as a single company's failure.
That framing depends on how the breach is understood. Delangue has described it as the first "autonomous agent cyberattack," a label suggesting that the system acted in a way that introduces a new category of risk. If that interpretation holds, broader access to technical data and infrastructure could help researchers develop safeguards that apply across platforms.
– clem (@ClementDelangue) July 28, 2026
Not everyone agrees with that characterization. Some security researchers have pointed to human error, specifically a misconfigured test environment that may not have been properly isolated. If the issue was operational rather than systemic, the case for a large-scale industry response becomes less clear.
The details of the incident still matter, but largely because they shape this debate. OpenAI said two of its models, including GPT-5.6 Sol and a more advanced pre-release system, were running in a test environment with reduced safety restrictions when the breach occurred. One of the agents obtained an access key and moved deeper into Hugging Face's network.
What followed reinforced Delangue's argument for greater openness. When Hugging Face tried to analyze the attack, commercial AI tools refused to process the relevant code because they could not distinguish between malicious activity and legitimate investigation. The company instead turned to GLM 5.2, an open model developed by Z.ai and running on its own systems. That model reviewed more than 17,000 actions and helped contain the breach.
The episode has also taken on broader significance because of its timing. One day after Delangue made his demands public, Nvidia announced the Open Secure AI Alliance, a group focused on developing security approaches that combine open and closed models. Hugging Face is a member; OpenAI is not. Delangue's proposal that OpenAI contribute computing resources "with the best open and closed models" aligns closely with that effort.
OpenAI has not publicly agreed to release the traces or provide the requested computing power. Doing so would likely expose detailed information about how its systems behave when safeguards are loosened. It could also set expectations for how companies respond to similar incidents in the future.
For now, Delangue's approach stands out as much as the incident itself. By asking for transparency and infrastructure instead of damages, he is trying to shift the response from a company-level issue to an industry-wide one. Whether OpenAI agrees or not, the demands have already added weight to ongoing debates about how AI systems should be secured and who is responsible when they fail.