US federal agencies quietly sidestep Trump’s Anthropic ban for Mythos
TL;DR
- Federal agencies and officials are working with Anthropic despite President Trump’s ban on the company, Politico reported on 14 April; Reuters relayed the story the following day
- The Commerce Department’s Center for AI Standards and Innovation is actively testing Mythos’ hacking capabilities, with congressional committee staff also seeking briefings
- The situation illustrates how national-security imperatives cut through political procurement blocks — a useful reference point for UK AI sovereignty debates
Federal agencies and government officials are quietly circumventing President Trump’s ban on working with Anthropic, according to a Politico report picked up by Reuters. The Commerce Department’s Center for AI Standards and Innovation is actively testing Anthropic’s frontier AI model, Mythos, for its cyber-offensive capabilities, Politico said. Staff on at least three congressional committees held or requested briefings from the company about Mythos’ cyber-scanning capability over the past week.
Security overrides politics
The context for the workaround is the Pentagon’s decision to cut off business with Anthropic following a contract dispute — a move Trump endorsed. Yet Mythos-class capabilities are too consequential for federal cyber teams to ignore. Anthropic co-founder Jack Clark said at the Semafor World Economy event on Monday that the company is still discussing Mythos with the Trump administration despite the Pentagon cut-off.
Reuters could not independently confirm the Politico report. Anthropic, the White House and Commerce Department did not respond to requests for comment. But the direction of travel is clear: when a frontier AI model demonstrates capabilities that reshape national cyber risk, national security machinery finds pragmatic ways to engage with it, political constraints notwithstanding.
The UK sovereignty mirror
The pattern is a useful comparator for the UK Sovereign AI Fund launching this week. UK Technology Secretary Liz Kendall’s £500m initiative aims to back domestic AI champions precisely because access to frontier-lab capabilities increasingly determines national security posture. The US situation — federal agencies contorting around a Presidential ban to maintain access to Anthropic’s latest model — underscores why. Frontier AI access is now a strategic resource, not a procurement preference. Countries without domestic capacity are exposed to precisely this kind of political whiplash.
It also illustrates the limits of procurement bans as AI policy tools. When the technology in question affects cyber defence readiness, exclusion creates more risk than it contains.
Looking forward
Expect the Anthropic-White House dynamic to evolve as Mythos scrutiny continues at the ECB, Bank of England and NCSC. The Pentagon block may prove politically untenable once federal cyber agencies publish their evaluations. For UK readers, the underlying lesson is the more durable one: owning frontier AI capability domestically is a hedge against the politics of any single trading partner’s AI supply chain. That argument sits behind the Sovereign AI Fund — and this week’s US story reinforces it.