U.S. intervention forces Anthropic to halt advanced AI

2026-06-13

Anthropic on the 12th halted the offering of its advanced AI model, Claude Mythos, after the U.S. government intervened under export control rules for security reasons to stop foreigners from using it. As AI grows more important in cyber defense, the clash between the company leading development and the U.S. government has come into the open.

Mythos and Fable halted

The company said the same day it would stop providing Mythos for cyber defense and Fable for general users. Both are AI models with top-tier industry performance. Mythos had been used by 200 companies and organizations in about 15 countries, including Japan and the United States, while Fable had been used by consumers and businesses. Lower-performance models such as Opus will continue to be offered.

According to the company, it received a letter from the government on the 12th ordering it to stop foreigners from using Mythos and Fable. The letter called for ending access for all foreigners in and outside the U.S. and for the company's foreign employees, and Anthropic said it responded by stopping access for customers worldwide, not just foreigners.

Unusual order to suspend services

It is unusual for the U.S. government to use export controls to order a company to suspend a model after it has already been released, given that models are the underlying technology of AI. As social risks from the misuse of advanced AI rise, the U.S. government has moved to take a more forceful approach. The fact that AI can also be adapted for military use, and is becoming a strategic resource for nations, is also behind the intervention.

Mythos is highly capable of finding system vulnerabilities, and a prolonged suspension could slow global cyber defense efforts. Governments, financial institutions and infrastructure companies had been using such high-performance AI to quickly fix flaws in existing systems.

It is not clear whether Mythos Preview, the previous generation, is also subject to the suspension. Japan's three megabanks and Hitachi had only just begun using the model. Finance Minister Satsuki Katayama posted on X, formerly Twitter, that 'at this point, there has been no change in the understanding reached between the Japanese and U.S. finance ministries.'

Post-release intervention sharpens the conflict

Anthropic announced Mythos in April, but because of its high performance it kept it closed to the general public on its own judgment, citing concerns about misuse for cyberattacks and other purposes. The Trump administration had traditionally prioritized promoting AI development and been cautious about regulation, but it has been revising its hands-off approach in response to the arrival of advanced AI.

After repeated talks with the U.S. government, the company expanded Mythos access to Japan and other countries on June 2. It added safety measures and made Fable public on the 9th, but this time the suspension was triggered after the government pointed out a loophole in Fable that could bypass safeguards and be misused for cyberattacks.

Anthropic argued that the abuse was not serious and that there were no safety problems, saying it did not agree with the government's judgment and that there was a misunderstanding. It also said it aimed to resume offering the service as soon as possible.

U.S. Defense Department also involved in export controls

Since April, Anthropic has been in frequent contact with Treasury Secretary Scott Bessent and other senior administration officials. According to people familiar with the matter, Fable also underwent a safety review by an agency under the U.S. Commerce Department before its release. But because AI is a technology that can also be adapted for military use, other agencies, including the U.S. Defense Department, are also involved in export controls.

The White House had previously been wary of foreign companies and others using Anthropic's advanced AI. With China and others in mind, officials appear concerned that U.S. advantages in the AI race could be eroded, and they may also want computing resources to be used preferentially on the U.S. side. U.S. President Donald Trump signed an executive order on the 2nd requiring the U.S. government to pre-screen the safety of advanced AI, but intervention after release was not expected even by the company.

Long-simmering tensions

Anthropic and the U.S. government have long had points of friction. A dispute over the scope of AI use for the U.S. military led the Defense Department in February to explicitly name the company as a supply-chain security risk. After negotiations broke down, Trump at one point said he would bar the company from the entire government.

The company sued the U.S. government in March, and the litigation is still ongoing. Some inside the administration and among investors close to Trump are wary of the company because it strongly advocates tighter regulation. For the fast-growing Anthropic, government intervention is a new business risk. If it is forced to change its services, it could drive customers away.

The company filed for an initial public offering on the 1st and is pressing ahead with the expansion of its enterprise AI business. Being unable to offer its most advanced models would be a setback for its business strategy. Gary Marcus, professor emeritus at New York University, said the decision to deny access even to foreign employees 'could stop Anthropic's activities and amounts to an overreaction by the government. It will work against the U.S. AI industry.'

Enjoyed this article? Share it with your network!