OpenAI has been briefing US federal agencies, state governments and Five Eyes intelligence partners on its latest cyber-focused AI model over the past week, as demand grows for advanced tools to strengthen digital defences.
According to Axios, the company recently demonstrated its GPT-5.4-Cyber model in Washington to around 50 cyber defence officials from across the US government and national security agencies. The model is being rolled out under a tiered access system through OpenAI’s Trusted Access for Cyber programme.
Officials who attended the briefing included practitioners responsible for day-to-day cyber operations. The model is currently being offered to selected government and commercial users following a vetting process similar to that used for private sector clients.
OpenAI is pursuing a dual-track strategy: one version of the model is being made more widely available with safeguards, while a more permissive version is restricted to vetted cybersecurity defenders.
The company is also engaging with state governments and expanding access discussions with Five Eyes allies — Australia, Canada, New Zealand and the United Kingdom — as part of early onboarding efforts.
At the Washington event, OpenAI officials said the aim is to help a wider range of organisations, including critical infrastructure operators, identify vulnerabilities more quickly, particularly in systems affected by legacy infrastructure.
The development comes amid parallel efforts by other AI firms, including Anthropic, which is also working with US agencies on cyber tools under restricted release programmes due to security concerns.
Government and industry users are increasingly exploring how advanced AI systems can assist in detecting vulnerabilities, analysing threats, and strengthening cyber resilience across both public and private sector networks.
Also read: Employees fear they are training their own AI replacements
For more videos and updates, check out our YouTube channel
