Anthropic PBC is nearing an agreement with the U.S. government that would lift export restrictions on its two most advanced artificial intelligence models, Fable 5 and Mythos 5, following several weeks of negotiations over security concerns. The discussions involve senior officials from the Trump administration, including Commerce Secretary Howard Lutnick, and representatives from Anthropic, notably co-founder Tom Brown.
The restrictions in question were imposed two weeks ago in a letter from Lutnick, which barred Anthropic from providing foreign nationals access to the models amid worries that the AI’s security safeguards could be circumvented. This move forced the company to suspend all international availability of the systems and sparked intensive talks aimed at resolving the issue.
According to individuals familiar with the negotiations, the Commerce Department is working toward alleviating security concerns to enable the removal of export controls, subject to approval from officials across the federal government. However, it remains unclear how long it will take for the White House and other agencies involved in shaping AI policy to grant their final endorsement. A U.S. official noted ongoing efforts to ensure alignment within the administration on any potential resolution.
Anthropic’s Fable 5 and Mythos 5 models have drawn significant attention due to their advanced capabilities. The company unveiled Mythos 5 to a select group of companies and institutions in April, cautioning that the model’s ability to detect cybersecurity vulnerabilities made broader distribution risky. Fable 5, a public-facing version of the Mythos-class model with additional guardrails, was released earlier this month, while the less restricted Mythos 5 was intended for limited security-focused deployment.
The government’s intervention represents one of the most assertive actions taken to date in regulating the commercial use of AI, reflecting growing concerns over the safety and control of such technology. The export restrictions intensified scrutiny on Anthropic just weeks after the company confidentially filed for an initial public offering with a valuation exceeding $900 billion.
Anthropic has positioned itself as a responsible AI developer but has faced several regulatory challenges recently. In addition to the Commerce Department’s export control measures, the company has been in conflict with the Pentagon over proposed guardrails for military applications of its AI tools. After contract negotiations broke down, the Defense Department labeled Anthropic a supply-chain risk in March and is seeking alternative AI providers.
In contrast to prior disputes, CEO Dario Amodei has taken a more conciliatory role in ongoing Commerce Department negotiations, reportedly reducing tension between the parties. Nonetheless, it remains uncertain what specific changes the government will require to address concerns over the possibility of “jailbreaking” or bypassing the safeguards on Fable 5. Anthropic has publicly criticized the export restrictions, arguing that enforcing a standard of perfect resistance to guardrail circumvention could effectively halt new model deployments across the industry. The company maintains that such absolute jailbreak protection is currently unachievable for any AI provider.
The government’s actions come amid a broader tightening of oversight on advanced AI systems. OpenAI recently announced limitations on access to a preview version of its GPT-5.6 model to approved government partners, citing similar concerns. As the U.S. intensifies efforts to regulate frontier AI technologies, the outcome of the Anthropic case could set important precedents for the industry going forward.