Anthropic disables Mythos and Fable model access

Anthropic says it will disable access to its most advanced AI models, Mythos and Fable, to comply with new US export-control requirements, according to Business Insider. The move lands as OpenAI has filed confidentially for a US stock market listing and as US cities such as Seattle weigh moratoriums on new datacenters amid power constraints.

Export controls are usually discussed in terms of chips and manufacturing tools, but software access is increasingly the choke point. If a model can be reached through an API from anywhere, a rule aimed at blocking foreign militaries or sanctioned entities has to be enforced by the vendor, account by account, region by region. That turns compliance into a product feature: geofencing, identity checks, usage monitoring, and the ability to cut customers off quickly. It also turns model training into a national-security-adjacent business where the regulator’s definition of “advanced” can change faster than a product cycle.

For Anthropic, the immediate economic effect is lost revenue from customers who were paying specifically for frontier capability, plus the cost of building and auditing controls strong enough to satisfy Washington. For rivals, it creates openings and pressures at the same time. Smaller labs outside the US can market “no US restrictions” access, while US firms that want to sell globally must price in the risk of sudden product withdrawal in certain markets. The burden falls unevenly: the biggest players can afford compliance teams and bespoke contract terms; smaller firms may find the export rule effectively a barrier to entry.

The policy also collides with the industry’s cost structure. Training and serving top-tier models already requires vast compute and power, pushing firms into long-term capacity deals and expensive infrastructure buildouts. When access is curtailed by government order, those fixed costs do not shrink with the customer base. Vendors still pay for reserved compute, staff, and datacenter commitments, while the addressable market becomes more conditional.

Business Insider reports Anthropic’s decision as a direct response to the Trump administration’s export-control regime. The practical test will be whether disabling Mythos and Fable becomes a one-off compliance action, or the start of recurring shutdowns that customers must treat as a normal commercial risk.

Anthropic is not changing a roadmap slide; it is switching off two named models that were already in the hands of paying users.