A call to the White House and access to Anthropic models

Events surrounding Anthropic and its flagship models Claude Fable 5 and Claude Mythos 5 have become one of the most high-profile episodes in the history of artificial intelligence regulation (we covered this here). According to several media reports, the trigger for the chain reaction was a phone call from Amazon CEO Andy Jassy to representatives of the White House administration.

The result of this communication was an unprecedented decision by the United States – the introduction of export controls on two commercial AI models. By Friday, foreign users lost access to them, and the cryptocurrency market reacted instantly with a surge in tokens linked to decentralized artificial intelligence.

How the crisis began
According to The Wall Street Journal, on June 11 Amazon’s research divisions identified a potential vulnerability in the Fable 5 model. During internal testing, the model was allegedly able to be prompted into generating information that could theoretically be used for cyberattacks. After this, Andy Jassy reportedly contacted senior officials in Washington. His warning was supported by at least five other major tech companies, amplifying the perception of a national security risk.

The White House launched an accelerated risk assessment process, including consideration of restricting access to the models outside the United States.

Negotiations with Anthropic and abrupt resolution
At the same time, US administration officials contacted Anthropic CEO Dario Amodei. According to Politico, they discussed a voluntary shutdown or temporary restriction of model access until the vulnerability was addressed.

However, according to the same reports, Amodei rejected the proposal, considering the threat assessment overstated. The company did not agree to a full market withdrawal of the model.

As a result, the White House decided to impose export restrictions, effectively blocking access to the models outside the United States.

Official positions
David Sacks, Chair of the President’s Council of Advisors on Science and Technology, said on X that the administration acted “reluctantly” and was “surprised” by the company’s refusal to comply with what he described as a standard vulnerability response measure. According to him, the restrictions will be lifted once the issue is fixed and model safety is confirmed.

Anthropic stated that the incident resulted from a misunderstanding and differing risk assessments. According to the company, the vulnerability was interpreted as more dangerous than it is in practice.

Amazon did not confirm details of its discussions with the government, stating only that engagement with regulators on safety issues is standard practice for major cloud providers.

Anthropic said it is working on restoring access for users. Industry estimates suggest Claude has around 18,900 monthly active users, making the incident significant but still niche in terms of mass market impact.

Market reaction: rise of decentralized AI
Financial markets reacted quickly and predictably. Restrictions on centralized AI models sent a strong signal to the decentralized technology sector.

Investors interpreted the incident as confirmation of political and regulatory risk surrounding large AI platforms. As a result, capital began flowing into projects positioning themselves as independent from government control.

Bittensor, often referred to as the “Bitcoin of decentralized AI,” rose nearly 24% in a single day. Venice Token (VVV), associated with the Venice AI platform, gained about 16%. Near Protocol, which develops infrastructure for AI agents and decentralized applications, posted gains of around 6%.

A regulatory precedent
Export controls on Anthropic models represent an important precedent. For the first time, commercial AI models were restricted internationally this quickly and decisively on security grounds.

This decision shows that regulators are beginning to treat advanced AI systems as dual-use technology, comparable in sensitivity to critical infrastructure. The further course of events will depend on how quickly Anthropic resolves the identified issues and demonstrates compliance with regulatory requirements.

Broader context
Historically, the situation resembles the Cold War era, when dual-use technologies were subject to strict export controls. At the time, CoCom played a key role in coordinating restrictions on strategic technology transfer.

The difference today is speed. Where approvals once took months, restrictions can now be imposed within a day. This reflects a new reality in which software models are becoming subject to near-instant government regulation.

Technical and infrastructure angle
The incident also highlights a structural issue in modern cloud AI systems.

According to several sources, geographic user segmentation proved more complex than expected. This may have contributed to the global rather than selective shutdown of access. If confirmed, it raises a fundamental question: can modern AI platforms provide precise, targeted access control without deep infrastructure redesign?

In other words, this is not only about model safety, but also about the architecture of the future AI market.

Conclusion
The Anthropic situation may mark a turning point in the relationship between governments, large tech companies, and the AI market. For the first time, it has become clear that access to frontier AI models can be restricted as quickly as access to critical technologies. Markets have already begun pricing in this new risk across assets – from equities to decentralized AI tokens.

A small visualization of Fable 5’s “glory” before the block is available on our Telegram channel