President Trump signed a sweeping National Security Presidential Memorandum on June 5, directing the US military and intelligence agencies to move fast on adopting commercial AI tools. The directive covers everything from battlefield targeting assistance to logistics and supply chain coordination inside defense operations.
The most consequential line in the memo is not about weapons. It states that no company, commercial or otherwise, can disable, alter, or degrade an AI system that US military personnel depend on without prior government approval. That is new territory. Until now, software companies could push updates, change terms of service, or pull products at will. This memo says: not anymore, once the government is depending on your tool.
This matters beyond defense contractors. Any company whose software might flow into a government supply chain, a government-linked logistics network, or a defense-adjacent industry is now in a different relationship with its own product. The government is not just a customer; it is a co-owner of continuity.
The memo also replaces the Biden-era AI policy for defense, which the White House described as having created a dangerous reliance on a single vendor. The new approach explicitly calls for AI from multiple suppliers: OpenAI, Google, Microsoft, AWS, Nvidia, SpaceX, Oracle, and a startup called Reflection AI all now hold classified Pentagon contracts, after Anthropic was effectively blacklisted in February for refusing to allow its tools to be used for what it described as mass domestic surveillance and fully autonomous weapons.
The Anthropic saga is worth understanding. The Pentagon wanted to use Claude, Anthropic's AI, for any legally permitted purpose. Anthropic drew a line at autonomous weapons and mass surveillance. The Pentagon declared it a national security supply chain risk, a label previously reserved for companies with ties to foreign adversaries. Anthropic sued. More than 30 researchers from rival companies, including Google's chief scientist, filed legal briefs in Anthropic's support. OpenAI stepped in and signed a deal days later.
A separate executive order signed on June 2 adds a pre-release review window. AI companies can voluntarily let the government test their most powerful new tools for up to 30 days before releasing them to the public. Originally drafted as a 90-day mandatory review, the window was shortened and made voluntary after direct lobbying by tech executives. The review process will be classified, run by the NSA, and focused on cybersecurity risks. Companies that opt out face no direct legal penalty, but experts point out that procurement rules, government contracts, and insurers can make voluntary standards feel required in practice.
For anyone running a business that sells to, partners with, or operates near US government contracts, the picture is this: AI tools are being woven into national security infrastructure at speed, the government is asserting new controls over how those tools can be changed or removed, and the companies that push back on those terms risk being shut out of the market entirely. The Anthropic case showed that even a principled stand, widely supported inside the industry, does not guarantee survival in government contracting. The memo signed this week makes that dynamic more formal and more permanent.