EU finally secures access to Anthropic’s Mythos AI after months of pressure and negotiations

The European Union is poised to gain access to Anthropic’s closely guarded Mythos artificial intelligence model, marking a significant breakthrough after months of discussions between the bloc, the AI startup and US authorities.

The development comes at a time when advanced AI systems are becoming increasingly intertwined with national security, cybersecurity and geopolitical competition. Mythos, widely regarded as one of the most capable AI models developed to date, has remained largely unavailable outside a small group of trusted organisations because of concerns over its potential misuse.

The European Commission confirmed that it has held several constructive discussions with Anthropic regarding future access to the model, suggesting that the long-running negotiations are beginning to bear fruit.

Why Mythos has remained under lock and key

When Anthropic unveiled Mythos earlier this year, the company deliberately chose not to release it broadly. Instead, access was restricted through its Project Glasswing programme, which allows a limited number of organisations to evaluate the model in controlled environments.

More from Tech

The decision was driven largely by fears that highly capable AI systems could be exploited for offensive cyber operations if released without sufficient safeguards. By limiting access, Anthropic aimed to ensure that vulnerabilities could be identified and addressed before the technology reached a wider audience.

Only around 40 organisations were initially invited to participate in the programme. These included major technology firms, cybersecurity specialists and infrastructure providers such as Amazon Web Services, Apple, Broadcom, Cisco, CrowdStrike, Google, JPMorganChase, the Linux Foundation, Microsoft, NVIDIA and Palo Alto Networks.

Later, the White House also opposed a proposal by Anthropic to expand access to the model, complicating plans to put the technology in the hands of a broader group of companies and organisations. But, now that seems to change.

The company’s strategy has been to place the technology in the hands of organisations responsible for securing critical software and digital infrastructure, enabling them to strengthen defences against emerging cyber threats.

Europe’s push for AI access

For months, European officials have argued that access to cutting-edge AI systems is essential if the bloc is to maintain technological competitiveness and strengthen its cybersecurity capabilities.

According to reports, discussions intensified in recent weeks after Anthropic informed European authorities that government approval would be required before the company could share the model more broadly. This reportedly prompted additional conversations between EU officials and the US administration.

The negotiations have unfolded against a backdrop of growing concern in Washington about sharing advanced AI capabilities with foreign governments. The United States has increasingly viewed leadership in artificial intelligence as a strategic advantage and has been cautious about granting access to some of the most sophisticated systems being developed by American companies.

Nevertheless, progress appears to have been made.

European Commission spokesperson for tech sovereignty Thomas Regnier welcomed the latest developments, telling CNBC, “We welcome the latest developments on potential future access.”

He also emphasised that the issue extends beyond a single AI system. “Let’s not forget that Mythos is not one off, a new wave of powerful models are coming to the market,” Regnier said.

“This is a shared challenge, and we are intensifying our discussions with like-minded partners, including the United States.”

The breakthrough follows another major AI access agreement secured by the European Commission earlier this year. In May, the bloc obtained access to OpenAI’s GPT-5.5 Cyber model, reflecting Europe’s broader effort to ensure it can evaluate and prepare for increasingly powerful AI technologies.

As AI capabilities continue to advance rapidly, the agreement with Anthropic underscores a growing consensus among governments and technology companies: safeguarding powerful models may require international cooperation as much as technological innovation.