AI Risk Clock
Doomsday Clock5 min to midnight
→
⚖️ Neutral edit
axios.com2026-04-22

Exclusive: OpenAI briefs feds and Five Eyes on new cyber product - Axios

ModelsFeaturesSafety

OpenAI has been briefing U.S. federal agencies, state governments, and Five Eyes intelligence allies on its new cybersecurity product, GPT-5.4-Cyber, over the past week. The company held an event on Tuesday in Washington, D.C., attended by approximately 50 cyber defense practitioners from across the federal government, where OpenAI demonstrated the model's capabilities. The product is designed to autonomously identify and exploit vulnerabilities in computer systems, ostensibly to help defenders improve security. Sources familiar with the briefings said the model achieved high success rates on benchmark penetration tests. OpenAI did not disclose plans for broader commercial availability or a timeline for release. The briefings precede potential procurement by government bodies. The Federal Bureau of Investigation was among the agencies represented. The product represents an escalation in AI capabilities applied to cybersecurity, following OpenAI's previous work on general-purpose models that can also be used for code analysis.

Read this story in another voice
● REC · 2026
Exclusive: OpenAI briefs feds and Five Eyes on new cyber product - Axios — Neutral edit · AI Risk Clock