What happened
TechCrunch reported on April 18 that Anthropic appears to be rebuilding ties with senior members of the Trump administration even after the Pentagon designated the company a supply-chain risk. The report cites earlier signs of a thaw, including accounts that Treasury Secretary Scott Bessent and Federal Reserve Chair Jerome Powell had encouraged major banks to test Anthropic’s new Mythos model.
The article then points to fresh confirmation from both sides. Axios reported that Anthropic CEO Dario Amodei met with Bessent and White House Chief of Staff Susie Wiles. The White House described the session as "productive and constructive" and said it covered collaboration and protocols for managing the challenges of scaling the technology. Anthropic likewise said it discussed shared priorities such as cybersecurity, America’s AI lead, and AI safety.
That stands in sharp contrast to the Pentagon dispute. TechCrunch notes that the supply-chain-risk label is typically reserved for foreign adversaries and could sharply constrain government use of Anthropic’s models. The company is challenging the designation in court, while the report says other parts of the administration may still want access to the technology.
Why this matters
This story matters because it shows AI access inside government is fragmenting into a power struggle rather than moving through a single coherent policy lane. One arm of the U.S. state can treat a frontier-model provider as a security problem while another treats the same company as a strategic asset worth briefing, testing, and possibly using.
That is a meaningful shift. Frontier labs are no longer just selling software to public agencies. They are becoming part of a broader contest over national competitiveness, cyber capability, and institutional control. Once that happens, procurement language stops being purely administrative and starts functioning like a lever in a much bigger political fight.
The strategic read
Anthropic’s position is especially revealing because the company has tried to build its public identity around safety and policy seriousness. If even that profile does not protect it from being tagged as a risk by one government actor while being courted by others, then the emerging market for official AI adoption will likely be turbulent, selective, and highly political.
It also suggests something else: leading model providers may have to manage Washington almost like they manage compute supply. Access to the state is becoming a strategic variable. The companies that navigate those relationships best may gain not only contracts, but also influence over standards, deployment norms, and what counts as acceptable frontier behavior.
Bottom line
Anthropic’s Washington thaw matters because it reveals that the battle for AI leadership now runs through government institutions as much as through model benchmarks. The question is no longer only who builds the strongest systems. It is also who remains politically usable when the state wants frontier AI on its side.
Source note
Source: TechCrunch, "Anthropic’s relationship with the Trump administration seems to be thawing," published April 18, 2026.