Trump memo accelerates frontier AI across national security agencies

Inside Cybersecurity reports that Trump’s June 5 national security memorandum gives 14 covered agencies and offices two jobs that usually move at different speeds: put frontier artificial intelligence models into the national security enterprise, and put governance policy and risk assurance on a 90- to 120-day clock. The director of national intelligence, Defense secretary and National Security Agency must develop a joint risk-management and assurance strategy within 120 days establishing baseline AI security practices. The Committee on National Security Systems and Office of Management and Budget director have 90 days to issue policy for AI use in national security systems, including implementation and reporting requirements.

The contractor signal is less polite than the fact-sheet language. Agency heads are directed to start a process for terminating contracts with companies that repeatedly demonstrate a pattern of conduct inconsistent with Section 2 of the memorandum. Waivers can last no longer than one year and are limited to mission needs such as operational imperatives, test and evaluation, threat intelligence sharing and other mission-critical applications, with risk mitigation and enhanced oversight. For vendors selling AI into Defense, intelligence, Homeland Security, Treasury, Energy or Justice channels, this is procurement policy with a termination lever, pending the strategy that defines the baseline.

The unresolved part is sequence. The memorandum demands acceleration now, then asks the same agencies to write the baseline practices inside 120 days and a classified annex inside 90 days. “Appropriate assurances” may be obvious in a White House fact sheet. They are less obvious in a contract file, model card, red-team report or authority-to-operate package when the annex could change the answer for particular threat models.

Other deadlines point in the same direction: within 90 days, Defense must update Directive 3000.09 on Autonomy in Weapon Systems; within 120 days, procurement processes are to be updated for onboarding advanced models from multiple vendors; and agencies are to develop private-sector partnerships to secure cutting-edge AI technologies, including against malicious distillation attacks. The memo also rescinds Biden-era National Security Memorandum-25. For practitioners, Monday’s work is mapping existing AI security, red-team, data-center, personnel-vetting and vendor-control evidence against requirements that are about to be formalized. Waiting for the 120-day strategy may be legally neat. It will be operationally late.