SAN FRANCISCO — OpenAI plans to significantly expand access to a cybersecurity initiative that gives vetted professionals access to advanced artificial intelligence models, as the company prepares for more powerful systems in the coming months.
In a blog post published Tuesday, OpenAI said it is scaling its Trusted Access for Cyber (TAC) program from a limited pilot to include thousands of verified individuals and hundreds of teams responsible for defending critical software systems.
The company said the expansion comes as it develops more capable AI models and seeks to strengthen defensive cybersecurity applications. As part of that effort, OpenAI said it is introducing a version of its latest model tailored for security work.
“In preparation for increasingly more capable models from OpenAI over the next few months, we are fine-tuning our models specifically to enable defensive cybersecurity use cases, starting today with a variant of GPT-5.4 trained to be cyber-permissive: GPT-5.4-Cyber,” OpenAI said in the post.
Balance Sought
OpenAI said the TAC program is designed to balance broader access to AI tools with safeguards against misuse, noting that such technologies can be used by both attackers and defenders. The company said its approach includes making tools widely available to qualified users, deploying systems cautiously and updating them as risks evolve, while also supporting cybersecurity professionals across the ecosystem.
To expand the program, OpenAI said it is introducing additional identity verification measures. The TAC program initially launched in February with automated verification, but the company said it is now adding new tiers of access for users willing to undergo more extensive authentication as legitimate cyber defenders.
According to OpenAI, individual users will be able to verify their identities directly, while organizations can request trusted access for entire teams. Existing TAC participants will also be able to apply for higher levels of access if they complete additional verification steps.
‘More Expansive Defenses’ to be Needed
“Over the long term, to ensure the ongoing sufficiency of AI safety in cybersecurity, we also expect the need for more expansive defenses for future models, whose capabilities will rapidly exceed even the best purpose-built models of today,” OpenAI said.
OpenAI first introduced the TAC program in February alongside the release of its GPT-5.3-Codex model, describing it at the time as a pilot aimed at accelerating cybersecurity defense research.
As the CU Daily reported earlier, AI company Anthropic announced April 7 that it has a program called Project Glasswing that allows select partners to gain early access to the upcoming Mythos model positioned for defensive cybersecurity work. The new Mythos AI is considered so powerful it led to an emergency meeting at the White House with major financial institutions because of its capacity to identify cyber vulnerabilities.
