What is Anthropic?

Before we dive deep into the recent developments that caused quite some discussions within the cybersecurity industry, let’s start with the basics. Anthropic is an artificial intelligence company, most known for its large language model Claude. Founded by former OpenAI leaders Dario Amodei and Daniela Amodei, the company focuses on safe AI development while specializing in Constitutional AI.

With the hype around AI continuing to become bigger and bigger, Anthropic was recently valued at almost 1 trillion USD.

Anthropic’s AI models Mythos 5 and Fable 5

In April of 2026, Anthropic launched its Mythos 5 model with the Fable 5 model following in June of the same year. Both AI models are based on the same Anthropic 5 model. Fable 5 can be best described as a “less powerful version” of the Mythos 5 model as it runs with a set of extra safeguards. Initially, Mythos 5 was withheld from a larger audience as Anthropic feared the model may be too advanced to exploit software vulnerabilities. Yet, in April 2026, Mythos 5 was released for a small group of trusted tech and cyber firms with the safeguards lifted in some areas.

What happened?

Reportedly, Amazon CEO Andy Jassy raised concerns about Anthropic’s security, which then led to the US government – under Dondald Trump – to take quick and drastic action. The government issued a so-called “export control directive” to suspend all access to Fable 5 and Mythos 5 by any foreign national (inside or outside the United States), even including foreign national Anthropic employees.

Consequently, Anthropic disabled all access to Fable 5 and Mythos 5 to ensure compliance. As the company stated, selectively blocking only foreign nationals would technically not be possible. Hence, both AI models were shut down worldwide.

Long story short: the Trump administration used a ‘kill switch’ to cut off foreign access to US AI systems.

Note: This does not affect access to other Anthropic AI models like Claude for example.

Reasoning of the US government

Reportedly, researchers had found a way to get through the guardrails of Fable 5, or “jailbreaking” it, which could pose a cybersecurity risk. The US government is specifically concerned that users could enter prompts to Anthropic’s AI model that would find a way to get around safeguards that were designed to avoid criminal acts like cyberattacks for example. The government fears that the AI could be deployed to intelligence services in countries like China or Russia. Yet, a public statement or specific details regarding the national security concerns were not given by the Trump administration.

Anthropic’s statement

For context, the recent directive is neither the first nor the only lowpoint in the strained relationship between Anthropic and US government officials. Their dispute started when the company refused the US military to use its AI models for mass surveillance or autonomous weapon systems. This then caused the US Defence Secretary Pete Hegseth to consider Anthropic a “supply chain risk” which could block Anthropic from doing business with the US government.

Regarding its most recent dispute, the AI company tries to calm things down, merely speaking of a ‘misunderstaning’. The statement, that was issued by the company, further reads that there are other AI models available that are capable of the same things the researchers were concerned of. The company also underlines its efforts to restore Fable and Mythos as soon as possible. Moreover, Anthropic states to “disagree that the finding of a narrow potential jailbreak should be the cause for recalling a commercial model deployed to hundreds of millions of people.”

In its statement, Anthropic further declared that the research behind the security concerns came from engineers at Amazon – an investor but also a significant competitior for Anthropic in the AI industry.

Note: At the time this blog post was written, access to Mythos 5 and Fable 5 still has not been restored.

What does this mean for the future?

AI will stay. And with that, regulations and security risks need to be talked about. The Trump administration stated national security concerns as the reason for shutting down Anthropic AI models. Seems like a legitimate concern, right?

If the US government continues what it has started – using a ‘kill switch’ – and other governments around the world follow its lead, then it seems we are entering a new era of tech sovereignty. An era in which the main question will not be how to regulate AI and how to avoid security risks but an era where as a country, as an economic region, you need to make sure that you have powerful tech tools available – even when foreign tools are being taken away from you.

USA and China dominate AI, as they control around 90% of global AI computing infrastructure. If ordering a ‘kill switch’ becomes the norm and does not stay an exception, then this can have serious consequences for the European digital infrastructure.

With its shut down, private businesses are now caught in a dispute between US tech companies and the US government, having no say in it at all. Continuing to operate your business while relying on US tech providers becomes continuously dangerous.

Turn ON Privacy in one click.

Get

European Digital Sovereignty

If this shut down has shown us one thing, then it is the vulnerability of Europe as it mostly relies on foreign providers, especially when it comes to AI. What would happen if another supplier decides to shut down over night? Which businesses could continue to run without it?

Therefore, the term ‘digital sovereignty’ and the trend behind it is gaining more meaning and momentum, especially nowadays. More and more people and countries realise that for too long they have relied on American tech and start to ditch Microsoft, for example.

Fortunately, it is also becoming easier to leave Big Tech and break free from its vendor lock-ins as more and more alternatives become available. Moreover, Anthropic’s shut down further underlines the importance of open source providers.

One of these European alternatives is Tuta, the company that builds quantum-safe encrypted email, calendar, and drive solutions. Not only is Tuta based in Germany but also open source and GDPR compliant.

Final thoughts

With an unprecedented move, the US government forced Anthropic to do something no other AI company had to do before: a global shutdown. That marks a new “milestone” in the relationship between AI companies and the US government. But what this “milestone” further highlights that Europe cannot afford it anymore to be dependent on American Big Tech. Europe must become technically and digitally sovereign. Especially from US-American AI models.

Luckily, there is a wide variety of European alternatives that show a world without relying on Big Tech is possible. Start your journey on digital sovereignty now and together we can built a private and secure web.