The Trump administration is quietly reversing course on Anthropic, the AI startup it spent months trying to squeeze out of government contracts. What began as pitched legal battles and supply-chain sanctions now looks like an awkward reconciliation, driven by a simple reality: the company's most powerful models are too useful to ignore.
The rupture started earlier this year when Pentagon officials and Anthropic couldn't agree on how the government could deploy the company's AI in classified military settings. Talks collapsed. The White House responded with escalating hostility, including an unprecedented move to label Anthropic a supply-chain risk, a designation typically reserved for foreign adversaries. At one point, officials even drafted an executive order designed to lock the company out of government systems entirely.
That strategy crumbled once Anthropic rolled out its Mythos model and federal agencies started requesting access to test it against other companies' advanced cyber tools. The Pentagon and Anthropic could continue feuding in court, but elsewhere in government, the demand for what the company could build proved irresistible.
Now the White House is working on executive action that could resolve the deadlock and establish new rules for how federal agencies use cutting-edge AI systems. No final guidance has been locked in, but tech firms and cybersecurity companies have been pulled into meetings at the administration to hash out the details.
The shift exposes a core tension in Washington's AI strategy. The Trump administration came into office committed to a hands-off, pro-innovation posture. But as models have grown more powerful and raised questions about national security and competitiveness, that light-touch approach has shattered. Government is now actively shaping policy around who gets access to the most advanced systems and how they're deployed.
Jessica Tillipman, associate dean for government procurement law at George Washington University, explained the problem this way: when one agency negotiates a contract that effectively becomes administration policy, other agencies that disagree have little recourse except to carve out their own exceptions. That's what has happened with Anthropic. The Pentagon's hard line created a vacuum that other parts of government rushed to fill.
The company has signaled willingness to work within government constraints. An Anthropic spokesperson said the firm is engaged in collaborative conversations with the administration on cybersecurity and America's competitive position in AI, and that compute capacity is not standing in the way of expanding access to Mythos among additional companies.
One major problem remains unresolved: the Pentagon shows little sign of softening. Defense Secretary Pete Hegseth used testimony on Capitol Hill to call Anthropic's leadership ideologically driven and questioned whether the company should have sole decision-making authority over government AI deployments. Whether any executive order can bridge that gap is still unclear.
Author James Rodriguez: "The White House needs Anthropic's technology but hates the politics of admitting it, so now we're watching them try to have it both ways with vague executive action while the Pentagon keeps the grudge alive."
Comments