The Trump administration is moving to strengthen coordination on artificial intelligence cybersecurity through a new industry clearinghouse to impr
ove the detection and patching of network vulnerabilities, especially in open-source software.
The clearinghouse, named Gold Eagle, was rolled out earlier this month by the Treasury Department, Department of Homeland Security and the Pentago
n, in consultation with AI companies, according to National Cyber Director Sean Cairncross, who briefed reporters on the initiative July 14.
It will allow for the exchange of information about potential vulnerabilities and is a product of the AI executive order President Donald Trump issued last m
onth. The administration is working with the Software Engineering Institute at Carnegie Mellon University to develop a platform that will service as the intake portion of the clearinghouse, allowing companies to voluntarily provide
their most advanced models for evaluation.
Separately, the Vulnerability Information and Coordination Environment will handle the dissemination aspect of the clearinghouse, allowing for the sec
ure sharing of information on vulnerabilities so that software defects can be fixed, Cairncross said. The clearinghouse will include open-source scanning, he added.
The initiative is a centerpiece of Trump’s June executive order that outlined his administration’s approach to addressing the cybersecurity threats raised
by AI. The directive followed months of wrangling over how much oversight the federal government should have over new powerful models, and offered a largely laissez
-faire approach. Trump stopped short of mandating safety tests on cutting-edge AI models and instead sought to make them available to the government with developers’ permission for a 30-day review period.
See more beautiful photo albums Here >>>
The administration, however, has taken steps that appear counter to that hands-off approach in recent weeks. Officials restricted Anthropic PBC from releasing their most advanced models over concerns they could be e
xploited by bad actors. And under pressure from the administration, rival company OpenAI separately limited the release of its GPT-5.6 model to government-approved partners.
Officials have since lifted foreign access curbs on Anthropic’s Claude Fable 5 and Mythos 5 models after the company created additional cybersecurity guardrails for it
s technology. Meanwhile, OpenAI rolled out its newest suite of GPT-5.6 models after weeks of discussion with government officials.
