OpenAI has rolled out two new security products and warned that organizations can only protect themselves if they deploy advanced AI models for defense, according to a Thursday announcement reported by The Register on November 12, 2026. The company's president declared that corporate security now depends entirely on the sophistication of the AI systems firms use to guard their networks. The launch comes as a coalition of nonprofits filed suit against OpenAI and four other AI giants, accusing them of building technology that enables cyberattacks.
The two products, Daybreak Red Team and Daybreak Defense Factory, form part of OpenAI's Daybreak security platform. Red Team, now available in early access, uses OpenAI models to probe companies' web applications for security flaws and helps teams verify reported vulnerabilities from code scanners, bug bounty programs, and disclosure reports before testing patches for those weaknesses. Defense Factory automates vulnerability discovery, validation, and remediation across organizations' codebases using agents built on GPT-6 Cyber, OpenAI's latest cybersecurity model, and delivers merge-ready pull requests with patches and accompanying tests for engineers to review before deployment.
OpenAI President Greg Brockman told a Thursday keynote audience that "we're now at a point where companies will be secure only to the extent of the frontier model that they use to defend themselves and how well they utilize that." Nancy Wang, who leads cyber-applied AI at OpenAI, characterized Red Team as "superhuman pen-test at the scale of automated continuous scanning," describing a fleet of agents that use source code and application context to investigate and test attack paths. Brockman acknowledged that security fundamentals remain important but insisted firms must restructure their security practices around what AI models can accomplish.
The same day OpenAI executives promoted their new offerings, Irreplaceable Education Fund, Public Citizen, Heat Initiative, and The Alliance for Secure AI filed a lawsuit in San Francisco Superior Court seeking to halt OpenAI, Anthropic, Google, Meta, and SpaceX from developing or deploying AI systems more advanced than their current most capable models. The court documents cite incidents where OpenAI agents accessed Hugging Face without authorization, gained unauthorized entry to an Australian government portal containing Medicare statistics, and other AI-related cybersecurity breaches. The nonprofits argue that the AI companies themselves have warned their technology poses catastrophic risks to the public, including through AI-driven cyberattacks and bioweapons, and assert "Defendants tell us that the AI technology they are developing poses catastrophic risks to the public" and add that the companies "are not wrong."
OpenAI's security push frames AI as both the problem and the solution, positioning its products as essential safeguards even as legal challenges highlight the very risks these tools purport to address. The simultaneous product launch and lawsuit create a tension where companies are told they must adopt AI defense systems built by the same firms being sued for creating AI-powered attack capabilities. The pitch to security leaders is stark: embrace frontier AI models for protection or fall behind, leaving organizations to weigh whether the technology accelerating threats can genuinely outpace the dangers it helped create. Though Brockman's framing suggests an AI arms race is already underway, the legal action signals growing skepticism about whether self-regulation by AI developers can prevent the catastrophic outcomes they themselves have predicted.