História
julho 1, 2026

U.S. to Allow Anthropic's 'Mythos' AI Model Release

The Trump administration has decided to allow Anthropic to release its Mythos AI model to over 100 U.S. companies and government agencies. The decision reverses a two-week export restriction that was put in place over national security concerns.

The Trump administration’s abrupt decision to re‑open Anthropic’s Mythos 5 AI model to more than 100 U.S. companies and agencies has become a test case for how quickly Washington will trade broad AI controls for selective, security‑framed access.

Liberal and conservative outlets agree on the basic facts: Commerce Secretary Howard Lutnick authorized limited access to the Claude Mythos 5 model for “trusted partners” after a two‑week shutdown triggered by export‑control and national security concerns. Both underscore that access to Anthropic’s companion model Fable 5 remains blocked, highlighting an unresolved rift between the company and the government over how far controls should go.

Where they diverge is in framing risk versus necessity. The liberal account, grounded in CNBC’s coverage, emphasizes the breadth of the government’s earlier order: Anthropic had to disable Fable 5 and Mythos 5 globally and cut off “any foreign national, whether inside or outside the United States, including foreign national Anthropic employees,” a move justified under “national security authorities.” It situates the clash amid a longer arc of friction, noting the Pentagon’s “supply chain risk” designation and ongoing litigation—suggesting the reversal is one tactical step in a larger, uneasy relationship.

The conservative Washington Examiner narrative leans into Mythos as a security asset, calling it Anthropic’s “most powerful AI model” and stressing that it is being redeployed to “a small group of cyber defenders and infrastructure providers.” It highlights administration claims of “significant progress” in talks and Anthropic’s pledge to work on “protocols and standards,” framing the deal as prudent calibration rather than a climb‑down.

Critically, both perspectives leave unresolved questions: whether the narrow U.S.-only reopening meaningfully mitigates foreign access risks, and whether concentrating advanced AI in the hands of a government‑screened club of “trusted partners” creates its own power and oversight problems.