OpenAI Warns Future AI Models Could Pose Cybersecurity Threats

OpenAI has issued a stark warning that its forthcoming AI models may possess the capability to generate functional zero-day exploits, targeting even the most secure computer systems. The technology could facilitate sophisticated cyberattacks on businesses or critical infrastructure, potentially causing real-world damage.

The rapid advancement is alarming. OpenAI's internal data shows performance on capture-the-flag security challenges surged from 27% with GPT-5 in August 2025 to 76% with GPT-5.1-Codex-Max by November 2025. This exponential improvement suggests each new model could reach what the company terms 'high' levels of cybersecurity capability—able to create working exploits for unknown vulnerabilities in hardened networks or assist complex intrusion campaigns.

While the Microsoft-backed firm is investing in defensive security applications, developing tools to help security teams audit code and patch vulnerabilities, the dual-use nature of this technology presents a fundamental challenge. The same knowledge that empowers defenders could equally benefit attackers. OpenAI acknowledges no single protective measure will suffice, requiring instead multiple layers of integrated security controls.