Anthropic and US govt officials hold talks as AI export controls remain in place

Amid a high-stakes dispute with the US government over export controls on its latest AI models, Anthropic’s senior executives reportedly met with Trump administration officials on Monday, June 15, but failed to reach a resolution on the issue.

Both sides are working to resolve things quickly, according to a report by The Information. It is the latest update in a series of events that were set in motion after Anthropic received an export control directive from the US Commerce Department on Friday, June 12, ordering the AI startup to suspend access to its latest AI models, Fable 5 and Mythos 5, for any foreign national, whether inside or outside the United States.”

Anthropic responded by disabling access to the two models for all its customers, including US citizens, in order to ensure compliance with the directive. The abrupt decision marks the latest flashpoint in Anthropic’s increasingly strained relationship with the US government following a clash with the US Department of Defense earlier this year.

It has been labelled as a supply chain risk, meaning that US defense contractors are prohibited from using the company’s AI technology as it purportedly poses a threat to US national security.

In India, the development has reignited a long-running debate over technological sovereignty with several Indian founders, investors, and policy experts discussing whether the country should accelerate efforts to build its own AI infrastructure, invest more aggressively in open-source alternatives, or continue relying on foreign frontier models.

How did Anthropic get here?

Last week, Anthropic unveiled two new foundational AI models called Fable 5 and Mythos 5, both of which are built on top of Claude Mythos Preview, which the company has described as a powerful model that excels at identifying hidden security flaws in software and thus, posing grave cybersecurity risks if accessed by bad actors.

Prior to introducing Fable 5 and Mythos 5, Anthropic said that it worked with the US government to test the models. While Fable 5 was released to paid subscribers with built-in guardrails, access to Mythos 5 was limited to organisations that are part of Anthropic’s ‘Project Glasswing’ cybersecurity initiative.

Story continues below this ad

However, Amazon CEO Andy Jassy and other tech leaders reportedly flagged concerns to senior Trump administration officials about security risks in Anthropic’s latest AI models.

“As a leading cloud provider that serves a large number of private and public sector customers, it’s not uncommon for governments to seek out counsel on potential security risks. When they occur, we don’t share the details of those discussions,” an Amazon spokesperson said in a statement.

At around 5:30pm ET on Friday, Anthropic received a formal letter from the US government that required the company to suspend the models. Anthropic reportedly did not receive any communication about its models posing a threat to US national security before Friday, June 12. A report by Financial Times said that the company was given 90 minutes to comply with the order.

Calling it a “misunderstanding”, Anthropic said that it believes the US government’s concern is around a “potential narrow, non-universal jailbreak,” where a user could bypass a cybersecurity guardrail and ask Fable 5 to “read a specific codebase and fix any software flaws.”

Story continues below this ad

“We disagree that the finding of a narrow potential jailbreak should be cause for recalling a commercial model deployed to hundreds of millions of people. If this standard was applied across the industry, we believe it would essentially halt all new model deployments for all frontier model providers,” it said.

What has transpired since?

Here is a brief sequence of events that have taken place since Anthropic’s two AI models were taken offline:

– David Sacks, a Silicon Valley investor and co-chair of President Trump’s Council of Advisors on ​Science and Technology, said that Anthropic CEO Dario Amodei refused to “fix the jailbreak or de-deploy the model” after “a highly credible trusted partner” reported it.

– News reports suggested that the White House imposed export controls on the Mythos-based models partly over suspicions that a China-linked group had accessed them.

Story continues below this ad

– US cybersecurity leaders urged the White House to lift the ban on Anthropic’s models, arguing that the move hurts defenders more than attackers.

– An Axios report, citing an anonymous Trump administration official, said that Anthropic struggles to communicate effectively with the government and had failed to honour a cybersecurity-focused executive order signed by President Trump.