OpenAI starts GPT-5.5 Cyber rollout for critical defenders through Trusted Access

OpenAI is keeping its most cyber-permissive model access gated.

On April 30, 2026, TechCrunch and Nextgov reported that OpenAI is beginning to roll out GPT-5.5 Cyber to “critical cyber defenders” through the company’s Trusted Access for Cyber program. OpenAI’s GPT-5.5 release page already positioned cyber-permissive access inside Trusted Access, and the April 14 program post introduced GPT-5.4-Cyber as the earlier restricted-access model.

What changed

The new signal is that GPT-5.5 Cyber is moving from system-card and program language into access rollout.

OpenAI is not making the model broadly available to all ChatGPT or API users. The company is using identity verification, trust signals, and stated-use review to decide who gets access. TechCrunch framed the move as notable because OpenAI had criticized Anthropic for limiting access to Mythos, Anthropic’s own cybersecurity-focused model.

Why it matters

Cyber models sit on a difficult boundary. The same capability that helps defenders reproduce and patch vulnerabilities can help attackers find, exploit, or automate them.

OpenAI’s approach is now closer to Anthropic’s than the public rhetoric suggested: powerful cyber assistance is available, but through controlled access rather than a standard public product tier.

Tool impact

For ChatGPT, this does not mean ordinary users get GPT-5.5 Cyber. The relevant product change is programmatic: high-trust security professionals may gain stronger defensive capability under a separate access process.

For Codex, the connection is stronger. OpenAI’s GPT-5.5 page says Trusted Access includes expanded cybersecurity capabilities through Codex, which makes secure account posture, repo permissions, and audit trails more important.

For Claude, the competitive framing matters. Anthropic’s Mythos restrictions no longer look like an outlier. Both major labs are converging on gated access for high-risk cyber assistance.

Buyer takeaway

Security teams should treat GPT-5.5 Cyber as a specialized program, not a ChatGPT tier feature. If you handle critical infrastructure, vulnerability research, or defensive operations, the practical next step is evaluating Trusted Access eligibility, internal approval workflows, and logging requirements.

What to watch

Watch whether OpenAI publishes clearer eligibility rules, acceptable-use boundaries, and evaluation data for GPT-5.5 Cyber. Also watch whether Anthropic responds with more Mythos transparency now that OpenAI is using a similar access pattern.