The US Department of Commerce has Removed the export controls that forced Anthropic The company on Tuesday said it will shut down its two most advanced artificial intelligence (AI) models – Fable 5 and Mythos 5 – for every user around the world, including India.
The reversal builds on an episode that began on June 12, when the Trump administration ordered Anthropic to refuse models to all foreign nationals on national security grounds. The order comes as Washington increases surveillance of advanced AI models amid concerns that they could be misused by military intelligence in China, Russia or other countries of concern.
Since the company could not reliably distinguish foreign nationals from other users in real time, it closed access to the models to everyone, including Americans.
The Fable 5, released in early June, is a locked-down public version of the Mythos 5, a model Anthropic had blocked from wide public release due to its unusually strong ability to identify vulnerabilities that could be exploited by hackers. Mythos 5 only went to a small group of companies and governments, which Anthropic calls Project Glasswing, allowing these select groups and entities to discover other potentially dangerous uses or vulnerabilities in their own infrastructure.
Anthropic was restored first limited access to mythos 5 to a small group of ‘trusted’ US organizations on June 26, before the Commerce Department lifted the ban on both models entirely on June 30.
However, the restrictions spread throughout the AI industry and raised questions about whether export controls could be used to regulate AI model access.
What happened?
According to Reuters, Anthropic said the US government export control order came after Amazon researchers discovered a way to bypass Fable 5’s security measures, allowing powerful AI models to identify software vulnerabilities or access other illegal information.
The company has now implemented a new security measure that prevents this behavior. Any blocked requests will be forwarded to its older Opus 4.8 model.
Also read: Chinese AI company Z.ai takes aim at Anthropic’s cloud mythos
Anthropic said that while this may be frustrating for users, the tradeoff was made to ensure that the Fable 5’s other capabilities remain widely available. The company also warned that it is “probably impossible” to make any AI model completely robust to “jailbreak”, a term used to refer to techniques that circumvent security measures.
“There will be many small jailbreaks, some narrowly harmful, and although no universal jailbreak has been discovered for the Fable 5 at the time of writing, expert security researchers continue to red-team it,” the company said.
Close coordination, oversight and what it means
Anthropic also said in a blog post that it is deepening its collaboration with the US government, giving designated government partners early access to both of its models.
A letter to Anthropic from US Commerce Secretary Howard Lutnick said the company has agreed to work diligently with the government on protocols for Mythos, Fable, and future models, and to notify the US government of any malicious activity. However, Lutnick said the department “reserves the right to re-evaluate the decisions made in this letter and reimpose the licensing requirement if circumstances change or Anthropic fails to comply with its commitments”.
Increased scrutiny of AI models Starting in June, when US President Donald Trump signed an executive order establishing a voluntary framework for AI developers to offer “covered frontier models” to the government for 30 days before releasing them to trusted partners.
Human rival OpenAI also faced sanctions. It said last week that it had delayed the full public launch of GPT-5.6 at the request of the US government, limiting access to a small group of verified partners.
The episode marks a shift in how Washington now views AI policies. Export controls have traditionally been used by the US for the hardware that powers these models, not for finished and publicly deployed software. Governments have withdrawn access to consumer products for some time, which suggests that this kind of interference could become a new normal in a post-AI world.
Also read: Why does Anthropic have an edge over OpenAI in the IPO race?
Isaac Harris, executive director of the Frontier Security Institute, a nonprofit focused on AI and national security, told Reuters there now appears to be a process for standards like the US model. But, he added, “there are still question marks over how equally dangerous capabilities coming from China with fewer guardrails will be handled by the administration in the US market.”
Anthropic vs Pentagon
shut-off was the latest human-us government A fight that had been brewing for months.
Dean Ball, a former Trump White House AI adviser and lead drafter of the administration’s 2025 AI Action Plan, said on a New York Times podcast that Anthropic’s relationship with the US government extends to the summer of 2024.
At the time, the Joe Biden-era Pentagon and Anthropic agreed to use the cloud in classified settings, including intelligence analysis, but with two carves: no domestic mass surveillance, and no fully autonomous lethal weapons, Ball said on The Ezra Klein Show.
Also read: What could the return of Cloud 5 mean and Anthropic’s long feud with the Trump government
The Trump administration extended the contract on the same terms through 2025, Ball said, before the relationship soured in the autumn and opened up in early 2026.
The decisive rift, he said, came when Under Secretary of War for Research and Engineering Emil Michel asked Anthropic to remove the clause prohibiting the use of the cloud to analyze bulk-collected commercial data, the capability at the center of the mass-surveillance objection.
Ball said the fight was unusually personal as well as political.
US Defense Secretary Pete Hegseth wrote on X that Anthropic’s “true purpose is unambiguous. To seize veto power over operational decisions of the United States military”, which was “unacceptable”. Trump described Anthropic as a “radical leftist, woke company” and its employees as “leftist mad jobbers.”
The matter came to a head in February this year, when Hegseth threatened to invoke the Defense Production Act and designate Anthropic a “supply chain risk.”
That label is typically reserved for technology considered too dangerous to sit anywhere in the U.S. military’s supply chain, and has been used against foreign companies like China’s Huawei over spying fears. This tag had never been imposed on any American company before.
After this Trump announced on Truth Social That every federal agency would stop using anthropic technology: “We don’t need it, we don’t want it, and we will never do business with them again!” The US President also warned the company to “be helpful during this phaseout period, otherwise I will use the full power of the presidency to force them into compliance, with serious civil and criminal consequences”.
Anthropic, whose Pentagon work is covered under the $200 million contract, said being designated a supply chain risk “would be legally inappropriate and would set a dangerous precedent”, adding: “Any threat or punishment from the War Department will not change our position on mass domestic surveillance or fully autonomous weapons.”
The Defense Department eventually signed an agreement with OpenAI, stating that it has the same red lines as Anthropic. OpenAI chief Sam Altman told staff in a note seen by the BBC that his company would reject military use “such as domestic surveillance and autonomous offensive weapons”.
Anthropic subsequently filed two federal lawsuits against the Department of Defense on March 9, alleging that the ‘supply chain risk’ label was improperly applied and amounts to retaliation for the company’s protected speech on AI security, a violation of its due process rights.
“The Constitution does not permit the government to use its vast power to punish a company for its protected speech. No federal statute authorizes the actions taken here,” Anthropic had argued, telling a court that the designation would cause “irreparable harm” and put “hundreds of millions, or even billions, of dollars in lost revenue” at risk.
The cases remain unresolved. A federal judge in San Francisco granted a preliminary injunction barring the cloud from enforcing the ban, but a DC appeals court denied Anthropic an emergency stay on April 8.
(With inputs from agencies; This article is an updated version of an article published in June)
