OpenAI Launches GPT-5.4-Cyber and Expands Trusted Access Program

OpenAI expanded its Trusted Access for Cyber (TAC) program on April 14, unveiling the GPT-5.4-Cyber model for verified defensive security professionals. The model grants thousands of vetted individual defenders and hundreds of teams additional capabilities including binary reverse-engineering. GPT-Rosalind for biological reasoning and codec upgrades also launched in the same window.