Trump Administration Scales Back AI Order, Creates Cybersecurity Clearinghouse
Summary
A revised executive order makes federal testing of frontier AI models voluntary, limits access to up to 30 days and rules out mandatory licensing. Treasury will lead a cybersecurity clearinghouse for vulnerability coordination.
Key points
- Companies may voluntarily provide federal access to frontier AI models for up to 30 days before release, down from 90 days in earlier drafts.
- The order says the program is not a mandatory licensing or permitting regime and gives companies influence over which models are tested.
- Treasury will lead an interagency cybersecurity clearinghouse where government, private-sector and critical-infrastructure participants can coordinate vulnerability scanning, discovery, validation and remediation.
- Treasury and other agencies will develop classified benchmarks to identify advanced cyber and hacking capabilities in AI models.
- The order requires federal testing and model access to follow confidentiality, cybersecurity, insider-risk and intellectual-property protections.
- Experts and Senator Mark Warner raised concerns about vague implementation and potential politicization of testing.
Article Details
- Event Type
- Revised executive order establishing voluntary federal testing of frontier AI models and an interagency cybersecurity clearinghouse.
- Impact
- AI companies may voluntarily provide pre-release access to frontier models for up to 30 days; the order says the program is not mandatory licensing or permitting. Treasury will lead a clearinghouse for voluntary coordination on software vulnerability scanning, discovery, validation, and remediation, while Treasury and other agencies will develop classified benchmarks. The article reports concerns about vague implementation and possible politicization.
People
David SacksThe article identifies him as a tech investor and AI czar who lobbied against language in a previous draft, arguing it would impose too much regulatory burden on U.S. businesses.Mark WarnerThe senator commented on the order, praising some provisions while criticizing the administration's retreat from federal scrutiny and warning that testing partnerships could be politicized.Ryan FedasiukThe American Enterprise Institute fellow commented that questions remained about frontier capability access, vulnerability discovery for critical infrastructure, and sharing with trusted partners.Samir JainThe Center for Democracy and Technology representative commented on the order's approach to AI cybersecurity risks and warned that implementation should not be used to punish companies for political or arbitrary reasons.