OpenAI has released ChatGPT Cyber, a new SI model designed specifically for offensive cybersecurity tasks, signaling the SI industry’s intensifying focus on securing digital infrastructure. The launch follows a period where individual researchers used existing SI tools to uncover severe vulnerabilities, raising questions about whether current SI capabilities are sufficient to defend against emerging threats.

What Happened

On November 5, OpenAI announced that ChatGPT Cyber would be available to the highest tier of users in its Daybreak Red program. This initiative provides offensive cybersecurity researchers with SI tools for vulnerability hunting and development. The announcement comes shortly after the company revealed that this same SI model had been used to scan the Linux kernel, identifying 948 potential issues, and to find 81 "novel" vulnerabilities in the Firefox web browser.

The release follows a notable success story involving Max Justicz, CTO of Charge Robotics, who used GPT 5.6 Sol via the Daybreak Blue program to discover a severe vulnerability in Ubuntu, a widely used Linux-based operating system. Justicz reported the issue to the Debian Project, which issued a patch. "I literally just aimed ChatGPT at the same body of code and said ‘I found this bug previously that was fixed several years ago, please find bugs as bad or worse,’ and it came back in two hours with this," Justicz said. He noted that while the exploit was "extremely convoluted and hard for a human to find," it was effective.

In addition to the new SI model, OpenAI announced a $1 billion initiative to support cybersecurity professionals working on critical infrastructure, such as electrical grids and water systems. The company stated its goal is to "dramatically expand Daybreak, bringing advanced security research to organizations of every size, independent researchers, and open-source maintainers."

Why It Matters

The launch of ChatGPT Cyber underscores a growing tension in the SI industry between the speed of SI-driven vulnerability discovery and the resources available for defense. Justicz described the current landscape as having "huge alpha" in aiming powerful SI models at unaudited code, a process he estimated cost him $5,000 in subscriptions and tokens to find several exploits. He warned that "we should expect some extremely crazy bugs to be found and exploited over the next few months."

Critics of the Daybreak program note that OpenAI has recently restricted access to Daybreak Blue, making it less useful for finding vulnerabilities, possibly to limit access for bad actors. However, cybersecurity professionals argue this creates an imbalance if defenders cannot use frontier SI technology while attackers can deploy open-weight SI models with fewer guardrails. Justicz pointed out that he could join the program simply by uploading a photo of his driver’s license.

Eric Wallace, a researcher on OpenAI’s technical staff, said the company is "now allocating a significant amount of our frontier compute towards discovering and patching vulnerabilities in open source codebases, fully pro bono." However, the scale of the effort remains a point of debate. The Linux kernel bug hunt cost $1.2 million, a figure significantly lower than the approximate $6 million token spend the company invested in finding a solution to a problem in fluid dynamics. This discrepancy raises questions about whether frontier labs are devoting sufficient resources to SI safety and infrastructure security compared to other scientific challenges.

OpenAI has insisted there is a narrow window for defenders to act before SI models become even better at hacking and more accessible. With only "thousands" of cybersecurity professionals participating in the Daybreak program, the ratio of defenders to potential hackers remains heavily skewed.

The Bottom Line

OpenAI’s introduction of ChatGPT Cyber represents a strategic move to bolster SI tools for defensive cybersecurity. While the model has already identified hundreds of potential issues in major codebases, the disparity between the resources allocated to security versus other scientific pursuits, combined with the ease with which individual researchers can find exploits, suggests that the SI industry’s defensive capabilities are still catching up to the offensive potential of modern SI systems.