After 2 AI Models Disabled, Trump Drops National Security Threat Label—But DPA Powers Remain

The legal standing of AI export controls just became more nuanced. After forcing Anthropic to disable its most advanced models, Trump now says the company is not a threat—but he has not ruled out using the Defense Production Act. For regulatory lawyers, the episode raises urgent questions about the limits of executive authority over AI and the future of foreign national access to frontier models.

U.S. President Donald Trump's assertion that he no longer views AI lab Anthropic as a national security threat—a reversal from mere days ago—caps a weeklong drama that tested the boundaries of executive power over frontier AI technologies. In an interview with “The Axios Show” published on June 19, 2026, Trump acknowledged that “a week ago, maybe” he considered Anthropic a threat, but now, after high-level meetings and a G7 summit encounter, the company appears to have dodged a potentially crippling designation. The episode, triggered by a White House order for Anthropic to block foreign nationals from its most advanced models—dubbed Fable 5 and Mythos 5—forced the company to take the nuclear option: it simply disabled access for all users. Now, with Trump’s public softening, the immediate crisis is resolved, but the lingering threat of the Defense Production Act (DPA) and the precedent set for AI export controls cast a long shadow over the industry.

“

The episode, triggered by a White House order for Anthropic to block foreign nationals from its most advanced models—dubbed Fable 5 and Mythos 5—forced the company to take the nuclear option: it simply disabled access for all users.

The sequence began last week (around June 12) when, according to multiple reports, the Trump administration ordered Anthropic to shut off non-U.S. users’ access to its state-of-the-art models on national security grounds. The company complied swiftly, but in doing so it also cut off domestic users, a blunt move that underscored the severity of the directive. The models, Fable 5 and Mythos 5, represent Anthropic’s most powerful offerings, alleged to be competitive with systems from OpenAI and Google DeepMind. Their abrupt disappearance from public access rattled enterprise customers, researchers, and developers who depend on them. The order was seen as an escalation in the U.S. government’s increasing use of export controls and other trade-related authorities to regulate AI, mirroring earlier controls on semiconductor exports to China.

Then came a flurry of diplomacy. Senior Anthropic technical staff were scheduled to meet with administration officials earlier this week to hash out the dispute. At the G7 summit in France, Trump and other world leaders met with tech bosses, including Anthropic CEO Dario Amodei. The direct engagement appears to have paid off. In the Friday Axios interview, Trump praised Amodei for responding “very quickly” and “responsibly” to the export control directive. The president’s choice of words signaled that Anthropic’s cooperation had defused the confrontation. While he did not fully close the door on invoking the Defense Production Act—a Cold War-era law that grants the president sweeping power to direct industrial production for national defense—he suggested he might not need to. “I have the power to use a lot of things,” Trump said, “But I’m not sure I have to do that.”

The episode has significant implications for the AI landscape. For Anthropic, the rapid de-escalation might be seen as a win: it avoided being officially labeled a national security threat, which could have jeopardized contracts, funding, and its global user base. The company’s statement—“We are grateful to the administration for their ongoing partnership”—strikes a conciliatory tone, aiming to project a cooperative relationship with the White House. However, the public revelation that the U.S. president was, even briefly, considering such a designation has already rattled investors and partners. The AI startup’s valuation and future fundraising could face headwinds if geopolitical risk becomes a permanent factor in its credit profile.

For the broader AI industry, the standoff marks a pivotal moment in the balance between innovation and national security. The use of export controls on AI models—rather than just hardware—is relatively uncharted territory. That the order forced Anthropic to disable models globally, even for Americans, reveals a bluntness in current regulatory tools. The potential invocation of the DPA could set a precedent allowing the president to commandeer AI model access, development, or even training runs in the name of defense. This looms large for other frontier labs like OpenAI, xAI, and Meta, all of which have global user bases and foreign talent.