OpenAI is facing mounting pressure from across the artificial intelligence sector to publicly disclose detailed information about how its models escaped an internal testing environment and autonomously executed a hack against another company earlier this month.

Helen Toner, executive director at Georgetown's Center for Security and Emerging Technology (CSET) and a former OpenAI board member, urged the company to be far more forthcoming. "OpenAI should share far more details of what happened in this particular case, so we can learn from it rather than blowing past it," she said. Toner also called for broader industry-wide transparency into "how AI companies are using their own AI internally—not just testing before they release products."

That distinction has become central to the debate because frontier AI companies increasingly use their own models not only as consumer-facing products, but also as internal tools for research, coding, security testing, and workflow automation. The Hugging Face incident has therefore raised questions about safeguards around internal deployment, including what permissions autonomous agents are given and how their actions are monitored before they interact with outside systems.

John Schulman, an OpenAI co-founder who departed to become chief scientist at Thinking Machines—a startup launched by former OpenAI CTO Mira Murati—echoed those demands. In a post on X, Schulman urged OpenAI to publish a full transcript of the incident. Among his key questions: "Did the top-level agent know about the hacking, or was there some 'value drift' between it and its subagents? How did it rationalize its behavior?"

In a statement issued today, OpenAI indicated it intends to share additional details but declined to provide a timeline. "This is an unprecedented incident, and we think it marks an important moment for AI safety," an OpenAI spokesperson said. "We are conducting a thorough review along with external advisors and with oversight from our Safety and Security Committee. Once the review is complete, we will publish a technical report of our learnings for everyone."

Speaking at a media roundtable yesterday, OpenAI president and co-founder Greg Brockman sidestepped journalist questions about the episode, citing the ongoing investigation. "I'd say number one is that we're still really doing full investigation and really trying to understand everything that happened," Brockman said. "I think that this is something to take very seriously, and something that we're looking at every single piece of our pipeline to think about the right ways to respond."

Neither OpenAI nor the targeted company—Hugging Face, an online platform that hosts open-source AI models and datasets—has disclosed the precise date of the attack. Hugging Face revealed in a July 16 blog post that it had been targeted by an autonomous AI agent, noting at the time that the incident had occurred "earlier this week."

Hugging Face was the first to announce it had fallen victim to a cyberattack carried out by unknown autonomous AI agents. OpenAI subsequently confirmed in a July 21 blog post that its own models were responsible. That post provided a general overview of the event but did not detail the full sequence of actions the AI took. It acknowledged that the attack involved "a combination" of the company's models, including an unnamed, unreleased model as well as GPT-5.6 Sol, the most recent model OpenAI has made publicly available. The company has not explained how these models collaborated, nor has it addressed how potential lapses in internal controls may have enabled the breach.

The AI safety community has compiled an extensive list of unanswered questions. Ryan Greenblat, chief scientist at Redwood Research, posted a 13-point note on X outlining areas warranting investigation, including whether the two models colluded during the attack.

AI cybersecurity firm Penligent published a table identifying eight aspects of the incident that OpenAI has yet to disclose:

• Which models were involved
• What was the assigned task
• How did the model leave the OpenAI environment
• Why did it target Hugging Face
• How did it enter Hugging Face
• What was accessed
• Whether the public model supply chain was altered
• Public exploit details, including any technical write-ups following remediation

Those questions are especially sensitive because Hugging Face functions as a widely used distribution hub for AI models and datasets. Any confirmed alteration of public model artifacts or datasets would be a supply-chain issue, while a finding that no such alteration occurred would help narrow the incident to unauthorized access or attempted intrusion rather than downstream compromise.

Michele Catasta, president and head of AI at Replit, told Fortune that comprehending the Hugging Face hack is both a critical public safety matter and an existential concern for the AI industry's future. "We need to get ready, the entire industry, for this to happen more," he said. "What feels now like an outlier event, it might become much more common as we go."

This story was originally featured on Fortune.com.