Anthropic grants EU access to Mythos, the powerful AI that autonomously found 10,000 zero-day flaws

Anthropic is finally giving in to the European Union's demands to allow the bloc access to its most advanced Mythos AI model. The European Commission has confirmed that it had several ‘productive meetings’ with the San Francisco-based AI startup regarding getting access to Mythos, CNBC reported.

Notably, ever since Anthropic unveiled Mythos a couple of months ago, there have been widespread concerns about the cybersecurity implications of the model. The company has so far refrained from releasing Mythos publicly due to some of these concerns, instead granting access to a select group of companies under its Project Glasswing initiative to help protect against cyberattacks.

EU to get access to Mythos:

After months of demands to get its hands on Mythos, the EU is finally getting access to the powerful AI model.

EU tech sovereignty spokesperson Thomas Regnier told CNBC in an emailed statement, “We welcome the latest developments on potential future access.”

“Let’s not forget that Mythos is not a one-off, a new wave of powerful models are coming to the market… This is a shared challenge, and we are intensifying our discussions with like-minded partners, including the United States,” Regnier added.

Reportedly, the EU had been having discussions with the US administration in the past week to secure access to Mythos after Anthropic told the Commission it needed government permission to share the model.

The US is said to be against sharing Mythos with foreign governments as a whole as it tries to maintain its dominance as the AI leader of the world.

Notably, the EU has already secured access to OpenAI's GPT-5.5 Cyber model in May, but the European Commission had been holding meetings with Anthropic to get access to its most powerful model.

What is Mythos?

Mythos is the most advanced AI model by Anthropic so far, which it says has reached a level of coding proficiency where it can surpass all but the most skilled humans at finding and exploiting software vulnerabilities.

The company also said that Mythos had already been used to find high-severity vulnerabilities in every major operating system and web browser. Anthropic said that its AI model had uncovered a 27-year-old vulnerability in OpenBSD that could allow attackers to remotely crash affected systems and a chain of vulnerabilities in the Linux kernel that could potentially give attackers complete control of a machine.

In an update it shared earlier in the month, Anthropic said that the AI model was used by its approximately 50 partners to find more than 10,000 high- and critical-severity vulnerabilities across some of the ‘most systemically important software in the world’.

Anthropic also used Mythos to scan more than 1,000 open-source software projects, which it said ‘collectively underpin much of the internet’. The company identified over 23,000 potential vulnerabilities in these projects, including more than 6,200 that it classified as high- or critical-severity.

Meanwhile, Mozilla said that it found and fixed 271 vulnerabilities in Firefox using Mythos, more than ten times the number it discovered while testing with Anthropic's Claude Opus 4.6 model.