Anthropic vs. the White House. The company denies the allegations

In an emergency export control directive published on June 12, the US Department of Commerce imposed strict restrictions on Anthropic.
Read also: The fate of the CPN package is decided. The Prime Minister spoke on fuel prices
The manufacturer's flagship, newest tools cannot be distributed to any foreigners – both in the United States and abroad, including foreign employees of the company itself. Due to the global and networked nature of its services, Anthropic was forced to take a drastic step: completely disabling both systems for all its existing customers.
Too powerful to share
The locked systems debuted on the market just three days earlier, on June 9, and were immediately hailed as a technological breakthrough. Fable 5, positioned as a direct competitor to GPT-5.5 from OpenAI and Gemini from Google, demonstrated unprecedented capabilities in programming and implementing complex, autonomous tasks.
However, the architecture of the Mythos class was even more controversial. During April lab tests, its preview version proved so effective at automatically detecting and exploiting vulnerabilities in operating systems that the manufacturer itself described it as “too powerful for public release.”
Read also: The tension is rising. Ukrainian summit in Poland, loan dispute, SpaceX debut
Although security filters have been introduced in the commercial version of Fable 5, Donald Trump's administration decided that the risk to cyberspace was too great. According to the official statement, the government obtained a report demonstrating a security bypass method that allowed the model to analyze code bases and identify critical vulnerabilities. This could be used by hostile hackers to attack critical infrastructure, such as energy or banking.
Silicon Valley denies the allegations
The technology's creators strongly deny these accusations. Anthropic emphasizes that Before its premiere, the system underwent thousands of hours of red-teaming tests with the participation of, among others, the British AI Safety Institute and US government agencies.
At the time, Fable 5 firewalls were shown to be the most effective in the industry, and the vulnerabilities mentioned by officials are widely known and replicable in competing models such as GPT-5.5. According to the company, the government took drastic measures based on one narrow and innocuous presentation, which constitutes a profound misunderstanding.




