Anthropic disabled access to its Fable 5 and Mythos 5 models for foreign nationals on June 12, 2026, following a US government export control directive, according to Anthropic's official notice and CNBC reporting.

The restriction targets individuals without US citizenship, regardless of where they work or what IP address they use. This is different from the traditional export controls that block access by country or region. Al Jazeera confirmed that the order applies to foreign nationals specifically, with no apparent country-by-country exceptions. That distinction matters because it means the control follows the person, not the location.

Fable 5 and Mythos 5 are Anthropic's most advanced AI models. When a directive names two cutting-edge production systems like this, it signals that the US government has formally classified them as controlled technology under export law — likely under the same framework that has governed advanced semiconductor exports since 2022 and 2023.

Enforcing this order will be harder than it sounds. Blocking access by geography is simple: you check where someone is logging in and say no. Blocking by nationality requires Anthropic to verify who users actually are at the account level, something most consumer or API platforms aren't built to do at scale. The company hasn't yet said whether it will ask users to attest to their nationality, shut down unverified accounts, or take some other approach.

For large companies — especially multinationals with teams spread worldwide — the practical impact will be immediate. If a foreign national employee tries to use Fable 5 through their US parent company's account, that violates the directive. Legal and procurement teams at affected firms will need to audit their current usage quickly to see if they're in compliance.

The US has been layering AI controls into its export restrictions for the past year. Previous administrations focused on controlling the chips and hardware needed to train AI models. This directive goes further: it restricts access to the actual models themselves, at the inference layer where users interact with them. That's a new intensity in how the government is regulating frontier AI.

Here's what makes this significant. The directive creates an uneven playing field for companies. Anthropic, a US-based lab, now faces compliance costs that Chinese, European, and open-source AI developers don't have to manage. Whether the administration sees this as necessary national security protection or an awkward fit of Cold War-era export law onto 21st-century technology will likely fuel debate in Congress and among businesses affected by the order.

For now, the models are offline for foreign nationals. The policy is in effect.