The White House wants frontier AI on cyber defence, and it wants it fast. Gold Eagle, a new AI-backed clearinghouse, will pool software vulnerability findings from government and industry. It ranks the worst, then coordinates fixes across US critical infrastructure.
The pitch for AI-written malware is that it moves at machine speed. Washington has now built something to answer it at the same pace.
On Tuesday the White House launched Gold Eagle, an initiative it calls a clearinghouse for cyber vulnerabilities. In a press release, the administration said the system would “receive and patch” flaws “at a speed and scale never seen before”.
The plan pulls together the White House, the Treasury, and Homeland Security through CISA. The newly renamed Department of War joins them. They are working with open-source software groups and critical-infrastructure companies that the government did not name.
A wartime footing for cyber #
The language is unusually martial. Secretary of War Pete Hegseth said Gold Eagle brings “a wartime footing to the cyber domain.” He called it “the vanguard of America’s cyber defence.”
Treasury Secretary Scott Bessent framed it around the financial system. He promised to “harness frontier AI capabilities to stay ahead of our adversaries.” The idea is to use the same models that can find flaws at scale to close them first.
Gold Eagle grows out of an executive order Trump signed on 2 June. That order pushed advanced AI developers to give the government early access to their systems. The clearinghouse, the administration says, has already started taking in and ranking reports.
Long on ambition, short on detail #
The announcement leaves a lot unsaid. As Nextgov/FCW reported, the release reads mostly as a coordination mechanism. It does not suggest the scheme can force any company to fix a flaw.
The White House also left the basics open. It did not name the agency running Gold Eagle each day. It did not say how it would protect sensitive vulnerability data.
There is also company. CISA already runs a disclosure programme, alongside the CVE system and NIST’s National Vulnerability Database. Gold Eagle joins a crowded field.
It also named no participating companies. But Anthropic is a likely one. The AI firm said last month it would give federal officials advance access to its threat-intelligence reports. It also agreed to join the clearinghouse created by the June order. Anthropic made that promise just after its export-control spat with the same administration, and did not comment.
The race the government is trying to win #
The timing is not a coincidence. Security researchers say AI now touches every stage of an attack. It turns a fresh disclosure into a working exploit within hours. OpenAI last week called GPT-5.6 its strongest cyber model yet.
Defenders working at human speed cannot keep up. So the administration has set a separate 1 August deadline. By then, the NSA and CISA must build a classified way to benchmark frontier models’ cyber powers. The process would flag the systems that need watching.
Gold Eagle is the offensive-tools argument turned inward. It points the same AI that speeds up attackers at the defenders instead. The launch does not answer the hardest question. Can a clearinghouse that cannot compel a single patch actually win that race?
Get the TNW newsletter #
Get the most important tech news in your inbox each week.