Anthropic Mythos Too Powerful for White House to Ignore
Get the Tech newsletter
Daily tech — startups, AI labs, chips, the launches that shape the next decade. Free.
- The White House is moving to reconcile with Anthropic after months of legal battles and Pentagon animosity, with meetings this week and no final guidance yet, per Axios sources familiar with the talks
- The Pentagon escalated the standoff by labeling Anthropic a 'supply chain risk' — a designation usually reserved for foreign adversaries — and weighed an executive order to bar it from government systems entirely
- Anthropic's Mythos model forced the reversal: the NSA began testing it alongside other frontier AI cyber models even as the Pentagon feud dragged on, per Axios
- Anthropic disputed a WSJ report that the government opposed expanding Mythos access over compute limits, saying 'Compute is not a constraint' and that talks to bring in additional partners are ongoing
- Defense Secretary Pete Hegseth on Thursday called Anthropic's leader 'an ideological lunatic who shouldn't have sole decision-making over what we do' during Capitol Hill testimony, even as the White House was pursuing a thaw
- The Pentagon on Friday announced agreements with seven other top AI companies for 'lawful operational use' on classified networks, leaving Anthropic excluded from that arrangement
- Jessica Tillipman of George Washington University told Axios the dispute shows how 'regulating by contract' gives whichever agency negotiated the deal de facto power over the administration's AI policy
Why it matters: The Pentagon's attempt to blacklist Anthropic through contract terms created a de facto AI policy that other agencies are now actively circumventing, per GWU's Jessica Tillipman. With the NSA already testing Mythos despite the feud, procurement fights — not legislation or executive action — are now setting the direction of US AI policy across agencies.
