The company says that advancing capabilities increase the risk of exploitation and outlines defensive measures.

OpenAI cautioned that its next-generation artificial intelligence (AI) models could pose “high” cybersecurity risks as their capabilities grow.  The warning reflects growing concern within the tech industry that advanced AI could be misused to breach systems or enable novel attacks.

In a blog post accompanying the alert, OpenAI said the models could develop working zero-day remote exploits against well-defended systems or facilitate complex enterprise or industrial criminal operations.

OpenAI’s statement underscores that these risks arise as underlying technologies learn to operate autonomously for longer periods and gain broader access to code and systems. “As capabilities advance,” the company said, it is “investing in strengthening models for defensive cybersecurity tasks and creating tools that enable defenders to more easily perform workflows such as auditing code and patching vulnerabilities.”

To address the threats, OpenAI is tightening safeguards, including access controls, infrastructure hardening, limits on outgoing data flows, and ongoing monitoring to prevent misuse. The firm is also forming a Frontier Risk Council comprised of cybersecurity experts, initially focused on cyber threats posed by evolving AI models.

OpenAI plans to offer tiered-access programs for qualified users in cyber defense, providing advanced capabilities while reducing the risk of misuse. The company’s broader security strategy also includes enhanced tools for defensive use and third-party collaboration.

The alert comes amid growing scrutiny of AI’s impact on security and safety, including previous warnings about other emerging risks and regulatory discussions in the United States and abroad.