OpenAI released a new cybersecurity model. Is that enough?

OpenAI has released a new cybersecurity-focused model, ChatGPT Cyber, for its most senior tier of participants in Daybreak Red, the arm of its Daybreak program that equips offensive security researchers with tools for vulnerability hunting and development. The launch lands amid growing scrutiny of how frontier labs are handling AI's dual role in both finding and exploiting digital infrastructure flaws.

The timing follows a vivid demonstration of that duality. Max Justicz, CTO and cofounder of Charge Robotics, used GPT 5.6 Sol — released in July — through the program's defender-facing tier, Daybreak Blue, to find a severe vulnerability in Ubuntu that could have let hackers take control of servers running the widely used Linux-based operating system. Justicz, who has a hobbyist's interest in security and had previously found a bug in the same codebase, said he simply told the model to hunt for issues as bad or worse than one he'd seen fixed years earlier, and it returned a result in two hours. He reported the issue to the Debian Project, which patched it; there is no known evidence of exploitation. Justicz said he spent roughly $5,000 on subscriptions and tokens to uncover several exploits, and warned that "there is currently huge alpha in taking some code or binary someone has not audited yet and just aiming a powerful AI model at it."

OpenAI has been running its own scans at scale. The company said ChatGPT Cyber was used to scan and prepare patches for Linux kernel vulnerabilities, flagging 948 potential issues, and that it found 81 "novel" vulnerabilities in Firefox. The company says it wants to "dramatically expand Daybreak" to reach organizations of every size, independent researchers, and open-source maintainers, and has announced a $1 billion initiative supporting cybersecurity professionals in the program who work on critical infrastructure such as the electrical grid and water systems. Eric Wallace of OpenAI's technical staff said the lab is now putting "a significant amount of our frontier compute" toward discovering and patching open-source vulnerabilities "fully pro bono."

But defenders say the math still doesn't work. OpenAI insists there is a narrow window for those who safeguard digital infrastructure to act before models get better at hacking and easier to reach, yet practitioners describe a mountain of new work that outstrips their resources — and note that only "thousands" of cybersecurity professionals participate in Daybreak, versus a far larger population of potential attackers. Justicz also observed that OpenAI appears to have recently "nerfed" Daybreak Blue, making it less capable at finding vulnerabilities. That cuts both ways: he joined by uploading a photo of his driver's license, so gating limits bad actors, but defenders argue they fight with one hand tied while adversaries deploy open-weight models with stripped guardrails.

The open questions for builders and operators are about balance and follow-through: whether the promised expansion of Daybreak and the $1 billion in support actually reach open-source maintainers and small infrastructure teams before attackers move faster, whether defensive tiers stay nerfed as offensive capability scales, and how much frontier compute OpenAI and its rivals will truly commit versus the size of the backlog they've helped create.

Watch for signs in the coming months of whether patch pipelines for the Linux kernel and Firefox keep pace with the flow of AI-generated findings, and whether other labs match OpenAI's disclosed compute commitments.