OpenAI Launches GPT-5.4-Cyber to Reinvent Cybersecurity Defense

OpenAI has unveiled GPT-5.4-Cyber, a cybersecurity-focused model designed for verified defenders, featuring reduced refusal boundaries and binary reverse engineering capabilities. This release expands its Trusted Access for Cyber programme significantly after Anthropic's restricted Mythos model announcement. OpenAI’s approach emphasizes broad access under verification, challenging restrictive models as the cybersecurity landscape evolves.