Anthropic grants EU access to Mythos cybersecurity model after U.S. government approval

Anthropic is extending access to its Mythos AI model to the European Union following approval from the U.S. government, according to statements from the European Commission confirmed Monday.

The EU sought access to Mythos over several months due to cybersecurity concerns after the model's April release. Anthropic told the Commission it required U.S. government permission before sharing access, according to people familiar with the matter. The U.S. government initially resisted sharing the model with non-U.S. governments as part of efforts to maintain AI dominance.

What is Mythos

Mythos was initially rolled out to a limited number of companies in April as part of Anthropic's Project Glasswing cybersecurity initiative. According to Anthropic, the model excels at identifying security flaws and weaknesses in software and is capable of uncovering thousands of previously unknown software vulnerabilities.

The model's release prompted concerns from governments, banks, and tech firms about potential misuse by malicious actors to accelerate cybercrime.

EU negotiations

The European Commission held "several productive meetings" with Anthropic over the access request, according to EU tech sovereignty spokesperson Thomas Regnier. The Commission had "four or five" meetings with Anthropic and stated agreements were at a "different stage" compared to its May access to OpenAI's GPT-5.5-Cyber model.

Official terms and conditions of the EU access deal have not been disclosed. Anthropic did not immediately respond to requests for comment.

Broader AI security landscape

"Let's not forget that Mythos is not one off, a new wave of powerful models are coming to the market," Regnier said in emailed statements. "This is a shared challenge, and we are intensifying our discussions with like-minded partners, including the United States."

The Trump administration announced agreements in May with Google DeepMind, Microsoft, and xAI that allow the government to assess AI models before public release.

What this means

The decision to grant EU access marks a shift in U.S. policy on sharing advanced cybersecurity AI models with allied governments. While the U.S. initially resisted sharing Mythos to maintain AI leadership, the approval suggests growing recognition that cybersecurity threats require international cooperation. However, the undisclosed terms may include restrictions on model usage or further distribution. The precedent set by this agreement could influence access decisions for future advanced AI systems with dual-use capabilities.