Anthropic: Mythos Expands to Europe with ENISA

Anthropic and ENISA: Mythos Set to Cross European Borders

The European Union Agency for Cybersecurity, ENISA, is currently in discussions with Anthropic to gain access to Mythos, an artificial intelligence model specialized in cybersecurity. Until now, no European institution has had the opportunity to use this exclusive tool.

Mythos has generated significant interest in the cybersecurity sector since its launch. Developed by Anthropic, this model has distinguished itself not only by its performance but also by its rarity. Indeed, it has only been accessible to a select group of organizations, excluding the majority of the general public and governments.

ENISA in Talks for Access to Mythos

According to the Financial Times, Anthropic has proposed to the European Union to grant access to its AI model, with ENISA being among the first potential candidates. If these negotiations succeed, it would be the first time Mythos extends beyond the United States and the United Kingdom.

These discussions are part of the Glasswing project, a coalition initiated by Anthropic that primarily includes American companies. Since April, these members have been using Mythos to detect vulnerabilities and strengthen their cybersecurity.

This opening towards Europe raises important questions regarding digital sovereignty, cybersecurity, and technological influence. Although nothing has been signed yet, a spokesperson for ENISA confirmed that discussions are ongoing, specifying that the terms are still to be defined.

The European Commission is also involved in these exchanges. European representatives recently traveled to San Francisco to meet with Anthropic and discuss potential access to Mythos.

Access So Far Very Limited

Interest in Mythos is particularly high given that its access has been very limited since its launch. Anthropic chose to reserve it for a small group of organizations deemed essential, such as Microsoft, Apple, JPMorgan, and CrowdStrike.

This caution is explained by the advanced capabilities of the model, which, according to Anthropic, could be misused for offensive purposes in the cyber domain. This prospect is concerning, especially in a context where cyberattacks are becoming increasingly sophisticated each year.

Anthropic has also collaborated with several U.S. government agencies to evaluate the system before its deployment. So far, only the United Kingdom has obtained official access outside of the United States.

The potential arrival of Mythos within ENISA could enable the European Union to better anticipate certain emerging threats. European officials emphasize the importance of understanding the risks associated with the most advanced AI systems.