While Anthropic recently restricted its Mythos model to a select few companies, OpenAI is taking the opposite approach by broadening access to its security-focused model for thousands of developers. OpenAI has now announced a major expansion of its Trusted Access for Cyber (TAC) program, scaling it to thousands of verified individual developers responsible for protecting critical software. Along with this expansion, the startup also announced GPT-5.4-Cyber, a variant of the GPT-5.4 model fine-tuned for defensive cybersecurity use cases.
Earlier this year, OpenAI first announced the TAC program, offering automated identity verification for individual users and limited access for selected organizations to more cyber-permissive models. With today’s expansion, the startup is adding more access tiers. Customers in the highest tier will get access to the new cybersecurity-focused GPT-5.4-Cyber model.
For now, OpenAI said GPT-5.4-Cyber will roll out in a limited and iterative manner to vetted security vendors, organizations, and researchers. Customers who are already part of TAC can further authenticate themselves as legitimate cyber defenders to get additional tiers of access, including access to the new GPT-5.4-Cyber.
OpenAI highlighted that the new model is designed to be more permissive for legitimate cybersecurity work, lowering refusal boundaries for trusted users handling defensive tasks. For example, this model has advanced capabilities such as binary reverse engineering, which can help security professionals inspect compiled software for malware, vulnerabilities, and overall security, even without access to source code.
Here"s how to apply for the TAC program:
- Individual users can verify their identity at chatgpt.com/cyber.
- Enterprises can request trusted access for their team through their OpenAI representative.
OpenAI also hinted that more powerful models are coming soon, which may perform better than purpose-built models like GPT-5.4-Cyber. It said:
"Over the long term, to ensure the ongoing sufficiency of AI safety in cybersecurity, the company also expects the need for more expansive defenses for future models, whose capabilities will rapidly exceed even the best purpose-built models of today."
Earlier this year, OpenAI launched a $10 million Cybersecurity Grant Program, which reached more than 1,000 open-source projects with Codex for Open Source security scanning and improved Codex Security, which has helped contribute to fixing more than 3,000 critical and high-severity vulnerabilities since launch.