Key conclusions
- The US government has ordered foreign users to be cut off from the Claude Fable 5 and Mythos 5 models due to cybersecurity concerns.
- Due to technical reasons, Anthropic had to make the radical decision to immediately block global access to the systems for all commercial customers.
- The British AI Security Institute has shown in tests that the system can successfully break through firewalls in 73 percent of cases.
The US Department of Commerce and national security authorities forced Anthropic to abruptly restrict the distribution of its latest artificial intelligence systems. The official reason for the decision is the risk of bypassing Claude Fable 5’s security measures, which could allow foreign hackers to uncontrolled take over IT systems. The official administrative order concerned only the prohibition on making the technology available to non-US entities. However, due to the cloud architecture and the inability to immediately implement selective geo-blocking, Anthropic had to take the drastic step of completely disabling the Fable 5 and Mythos 5 models for everyone customers around the world.
The government’s decision caused immediate resistance from the manufacturer.
Anthropic representatives reported that the state administration did not present specific, unique vulnerabilities of the system. Officials cited the general discovery of a jailbreaking method that allows people to bypass factory security filters. The software developers verified the submitted attack demonstration and announced that the identified bugs were simple, already known, and could also be reproduced on competing models that are still publicly available without any restrictions.
Anthropic denies the Pentagon’s allegations and fights in court
The blocking of Claude Fable 5 is another stage of the escalation of the conflict between Donald Trump’s government and the modern technology sector. The conflict escalated when then-Secretary of Defense Pete Hegseth continued Anthropic status as a threat to the supply chain. This is the first time in US history that an American technology company received a security label previously reserved only for entities from geopolitically hostile countries.
This means a complete ban on implementing the company’s solutions in federal structures.
Anthropic took the case to court, accusing the Pentagon of acting without legal basis. A federal judge issued a temporary injunction, suspending the enforcement of the Defense Department’s directive. Thanks to this, government agencies and military contractors can formally use Claude systems, but the latest decision by the Department of Commerce to block the export of Fable 5 de facto prevents the exploitation of the software.
The British AI Security Institute reveals the model’s powerful capabilities
The controversy surrounding Claude Fable 5 results directly from its above-average operational capabilities. Before the June premiere, Anthropic conducted closed tests, making the system available to selected research organizations. The company itself admitted in official marketing messages that it had created a model that was too powerful to be safely released on the market, which led to accusations of aggressive marketing.
However, the numbers show the real scale of the problem.
The UK National Institute for AI Security conducted independent audits of the new software. The official report of this institution shows that Claude Fable 5 can independently identify, use and break the security of IT systems in 73 percent of attempts. Professor Gina Neff from Queen Mary University of London points out that such high effectiveness means a complete change in the rules of the game in cybersecurity and presents governments with challenges for which there are no procedures.
Europe reacts to US decisions and demands technological independence
The sudden blocking of access to American models caused an immediate reaction in Brussels. The European Union negotiated access to the earlier Mythos model in early June, after weeks of bilateral talks. European Commission spokesman Thomas Regnier said that officials are analyzing the situation and the whole event proves the need to build European technological sovereignty.
The EU has already implemented special regulatory and financial packages. Their goal is to radically make the economies of 27 Member States independent of artificial intelligence systems and server infrastructure from the USA and Asia. Experts from the London School of Economics note that cutting off European companies from American innovations may slow down the development of the local digital sector.
Recent events expose the main shortcoming of centralized systems – the state administration can paralyze them with one administrative decision. While Anthropic had to immediately shut down its servers, global blockchain networks prove fully censorship-resistant due to their distributed nature.
