OpenAI expands its Trusted Access for Cybersecurity program, granting global security experts access to the specialized GPT-5.4-Cyber model. The tool is fine-tuned for defense, featuring binary reverse engineering for malware analysis and lowered rejection thresholds for sensitive tasks. OpenAI's strategy emphasizes ecosystem resilience, highlighting its Codex Security tool which has fixed over 3000 high-severity vulnerabilities. The company states defensive AI capabilities must evolve alongside attack risks, deploying the model cautiously to approved researchers and vendors.