Why Trump Made America’s AI A National Security Weapon

Rows of illuminated server racks in a dark data center

The Trump administration is now deciding who gets access to the most powerful artificial intelligence tools in the world — and it’s keeping them out of foreign hands for national security reasons.

Story Snapshot

OpenAI launched a US-only preview of its new GPT-5.6 model series at the direct request of the US government, limiting access to just 20 trusted partners.
President Trump signed an executive order on June 2, 2026, creating a voluntary review process for powerful AI models before they reach the public.
Two weeks earlier, the government ordered rival AI company Anthropic to cut off all foreign users from its most advanced models, citing national security concerns.
The concern centers on “jailbreaking” — when someone tricks an AI into helping find security holes in critical computer systems.

GPT-5.6 Gets a Controlled US-Only Launch

OpenAI launched its new GPT-5.6 model family on June 26, 2026, but only inside the United States. The rollout covers three models — Sol, Terra, and Luna — with Sol serving as the flagship. Access is limited to 20 approved partners, with one path running through Amazon’s cloud service. OpenAI briefed the US government on the model’s capabilities before launch and shared the identities of its trusted partners with authorities.^[5]

OpenAI says it agreed to the restricted launch at the government’s request. The company acknowledged it is “uncomfortable” with the process and does not see it as a long-term solution. Still, it cooperated, recognizing the serious cybersecurity questions raised by a model this powerful.^[6] The GPT-5.6 family scores 91.9% on a leading cybersecurity benchmark, putting it directly in competition with Anthropic’s most advanced models.^[3]

Trump Signs Executive Order on AI Security

On June 2, 2026, President Trump signed Executive Order 14409, titled “Promoting Advanced Artificial Intelligence Innovation and Security.” The order tells government agencies to build a voluntary system for reviewing powerful AI models before release. It also directs the National Security Agency (NSA) to create a classified process for identifying which AI models are advanced enough to pose a national security risk.^[14]

Under the order, AI developers can give the government up to 30 days of early access to review a model before it goes to trusted partners. The order is voluntary — it does not create a mandatory approval process or licensing requirement. It also directs the Justice Department to make AI-enabled cyberattacks a top criminal enforcement priority.^[13]

Anthropic Was Ordered to Cut Off Foreign Users First

Two weeks before the OpenAI launch, the US Department of Commerce ordered Anthropic to block all foreign users from its top AI models, Fable 5 and Mythos 5. The government cited a reported “jailbreak” — a method that could trick the AI into identifying weaknesses in critical infrastructure. Anthropic pushed back hard, saying the evidence was based on a narrow, limited exploit that did not affect all users.^[7]

Anthropic also said it could not reliably enforce a ban on foreign nationals, raising real questions about whether these restrictions can even work in practice.^[1] The company temporarily disabled both models for all users — including Americans — while it challenged the order. Outside analysts noted that any exploit appeared limited in scope and that the models had multiple layers of protection already in place.^[10]

America First in AI: Smart Policy or Overreach?

The core concern driving these moves is real. Advanced AI models can scan software code and find security holes faster than any human. If adversaries like China, Russia, or Iran got unrestricted access to tools like GPT-5.6, they could use them to attack American infrastructure. The government’s push to control access before a full public release is a reasonable first step toward protecting national security in a fast-moving technology race.

Something historic just happened in AI governance and most people are framing it wrong.

The Trump administration asked OpenAI to restrict GPT-5.6 to a limited partner preview before broader release. Howard Lutnick personally warned Sam Altman not to launch without agency…

— Prasenjit Sarkar (@stretchcloud) June 26, 2026

That said, OpenAI’s own statement points out a tension worth watching. The company notes that cyber defenders — the people who protect American systems — also need these tools. Locking them out could slow down defense just as much as offense.^[6] The framework is voluntary for now, but the Anthropic case shows the government is willing to use direct orders when it feels the threat is serious enough. How Washington balances innovation with security will shape who wins the global AI race.