Introducing Trusted Access for Cyber

Source: OpenAI Date: 2026-02-05 URL: https://openai.com/index/trusted-access-for-cyber

Summary

OpenAI’s February 2026 launch of Trusted Access for Cyber — a controlled access program giving vetted cybersecurity professionals access to OpenAI’s most capable models for offensive security research, penetration testing, and defensive security work. The program addressed the dual-use tension in cybersecurity AI: the same capabilities that help defenders analyze malware and test systems also help attackers. Trusted Access created a gatekeeping mechanism allowing legitimate security professionals to use full model capabilities while restricting the same access from unverified users.

Implications

Credentialed access as a safety mechanism. Trusted Access for Cyber was a meaningful policy innovation: rather than restricting all security-relevant queries, OpenAI created a verified tier for security professionals with legitimate need. This required identity verification and use-case attestation, creating accountability for the access granted.

Thread: Agentic AI security and responsible deployment. Sits alongside the prompt injection safety post (January 2026), the April 2026 “Trusted Access for the Next Era of Cyber Defense” expansion, and the cybersecurity in the Intelligence Age piece (April 2026) as OpenAI’s coordinated cybersecurity product and policy strategy.

Watch: How the verification process for Trusted Access worked, whether security researchers found the program genuinely useful or bureaucratically onerous, and whether it influenced competing AI providers to create similar credentialed access programs.