OpenAI has unveiled “GPT-5.4-Cyber,” a specialised version of its latest AI models tailored for defensive cybersecurity applications. The announcement follows a recent move by rival Anthropic, which introduced its advanced model “Mythos” as part of its controlled-access initiative.
Anthropic revealed “Mythos” on 7 April 2026 under its Glass Wing project, a supervised programme allowing selected organisations to access a preview version of its model, “Claude Mythos,” for defensive cybersecurity purposes. The model has reportedly identified thousands of critical vulnerabilities across operating systems, web browsers, and other software.
OpenAI, the developer of the widely used chatbot ChatGPT, stated that GPT-5.4-Cyber will initially be deployed on a limited scale to vetted security providers, organisations, and researchers. The company is also expanding its “Trusted Access for Cybersecurity” programme to include thousands of individuals and hundreds of teams responsible for protecting critical software systems.
In parallel, OpenAI is introducing new tiers within its “TAC” (Trusted Access Controls) programme, launched in February 2026. Higher verification levels will grant users expanded capabilities, including access to GPT-5.4-Cyber.
Approved users at the highest tier will be able to utilise the model with fewer restrictions on sensitive cybersecurity tasks, such as vulnerability research and advanced threat analysis.
