OpenAI is rolling out GPT-5.5 and a specialized GPT-5.5-Cyber model to help cybersecurity defenders protect critical infrastructure. The company’s Trusted Access for Cyber (TAC) framework verifies defenders’ identities, reducing unnecessary refusals for tasks like vulnerability analysis, malware reverse engineering, and patch validation—while still blocking malicious uses like credential theft or exploitation. GPT-5.5-Cyber, in limited preview, is tuned for high-stakes protective work, but GPT-5.5 with TAC remains the strongest broadly useful defensive tool. Defenders must use phishing-resistant authentication starting June 2026. This move shifts from reactive security to proactive, AI-powered defense by putting powerful tools in the right hands. Why it matters: This is a blueprint for how frontier AI can be safely unleashed in high-risk domains—balancing capability and control without slowing down the good guys.