Trump Administration Imposes Export Controls on Anthropic's Most Powerful AI Models
N.R. Finch
The Commerce Department has placed Mythos 5 and Fable 5 under export control — the first time Washington has applied the regime to a specific commercial AI model — effectively treating frontier AI as a national-security asset.
What exactly happened?
Commerce Secretary Howard Lutnick sent a letter to Anthropic CEO Dario Amodei last Friday imposing export controls on Mythos 5 and Fable 5.
The controls cover three pathways: export, re-export, and in-country transfer. Each requires a license; Anthropic must also file a separate verification-license application.
This means → until licenses are granted, only U.S. citizens and permanent residents can access the two models, anywhere in the world.
Violations carry financial and civil penalties.
Why did the government act now?
The direct trigger: another company claimed it had successfully "jailbroken" the Mythos model — bypassing safety guardrails to extract outputs the model was not supposed to produce — raising national-security alarms.
The government had first tried to persuade Anthropic to delay releasing the two models. It failed, and only then turned to export controls.
In plain terms = Washington tried the soft block first; when that didn't work, it reached for the hard tool.
How awkward is Anthropic's position?
Anthropic now faces pressure from two directions: a Pentagon blacklist deems its models too risky for the government's own use, while a Commerce license regime deems them too risky for foreign use.
This reflects a rare situation — the same company's models are simultaneously judged "too dangerous for us" and "too dangerous for them."
The controls are expected to stay in place until the U.S. national-security apparatus is hardened — officials say that could take a matter of weeks.
Does this contradict the White House's own AI policy?
Earlier this month the White House issued an executive order requiring pre-deployment testing of frontier models, but it was explicitly voluntary and deliberately avoided a licensing regime.
Chief White House AI adviser David Sacks had argued against imposing "regulatory capture" on major labs — the risk that regulation ends up shielding incumbents rather than protecting the public.
This means → the Commerce letter effectively bypassed the White House's freshly laid down "voluntary, no-license" principle, using national security as the key to unlock mandatory controls.
One government official tried to bridge the gap: Trump "does not want to hurt the industry and wants innovation to keep moving forward."
What does this signal for the AI industry?
This is Washington's first export-control action targeting a specific commercial AI model. Previous controls covered chips and equipment — never the model itself.
Anthropic already had a pre-deployment testing partnership with Commerce's Center for AI Standards and Innovation (CAIS) — but that cooperation did not prevent unilateral action.
This reflects a clear signal: voluntary safety collaboration does not buy you a control exemption — cooperation is cooperation, and controls are controls.
Content is for reference only, not financial advice.