The AI startup held an event in Washington D.C. Tuesday (April 21) where it demonstrated its new GPT-5.4-Cyber model, Axios reported.
The report, citing a source familiar with the matter, said attendees included officials from throughout the government and from various national security agencies, most of whom are in charge of day-to-day cyber tasks.
OpenAI is also working with state governments to get them access to GPT-5.4-Cyber, the report said, and is beginning to brief the “Five Eyes,” a multi-national intelligence-sharing alliance made up of the U.S., Canada, the U.K., Australia and New Zealand.
The company is taking a dual-track approach, the report said, making one version of its model more widely available with robust safeguards, and another more permissive version for cyber defenders through its Trusted Access program.
OpenAI Chief Global Affairs Officer Chris Lehane said during Tuesday’s meeting that this tactic will let more companies, like local water utilities, access advanced AI tools.
Advertisement: Scroll to Continue
Sasha Baker, who heads the company’s national security policy, told attendees that OpenAI hopes to work with government departments to prioritize the most crucial usages and share threat intelligence across sectors.
This effort is happening soon after OpenAI rival Anthropic began previewing its Mythos AI model. That startup has held off on a wide release of Mythos, claiming it is too dangerous. Instead, Anthropic has offered the model to around 40 companies and organizations, including some within the government.
The report came the same day that Anthropic said it was investigating reports that a small group of people had gained access to Mythos through a third-party vendor. The company told PYMNTS it had found no evidence this activity extended beyond that vendor.
Meanwhile, PYMNTS wrote last week about an evaluation of Mythos by the U.K. Government’s AI Security Institute (AISI).
The chief takeaway from those findings, the report said, is not that AI can already carry out flawless cyberattacks. In fact, the AISI report noted that the success rate is limited.
“But systems that can plan and execute multistage intrusions, even inconsistently, represent a baseline that will improve,” PYMNTS added.
“More compute, better orchestration, and tighter integration with external tools will incrementally close the gap between partial and reliable capability.”
For all PYMNTS AI coverage, subscribe to the daily AI Newsletter.
