BNP Paribas and Mistral AI: Towards a Sovereign Cybersecurity Model

The collaboration between BNP Paribas and the French company Mistral AI marks an important step in the development of an artificial intelligence model dedicated to cybersecurity. This partnership comes at a time when Anthropic, the creator of Mythos, refuses to share this powerful tool with European entities.

Mythos: a powerful yet exclusive tool

Launched by Anthropic, Mythos is an AI model capable of detecting vulnerabilities and generating exploits for these flaws. In tests, Mythos managed to produce functional exploits on the first attempt in over 83% of cases, sometimes outperforming specialized human teams.

However, access to Mythos is limited to about forty organizations, primarily American, such as JPMorgan Chase. No European bank has been included, prompting institutions like BNP Paribas to seek alternatives.

Unsuccessful negotiations with Anthropic

Since April, the European Commission has been trying to gain access to Mythos, but discussions have not yielded results. Spanish officials have even described the situation as stalled. Anthropic fears that the dissemination of Mythos could turn it into a potential weapon in the event of misuse.

Mistral AI: a French alternative

In light of these obstacles, Mistral AI has been developing a cybersecurity model since May, presented as a sovereign alternative to Mythos. BNP Paribas has committed to this project, illustrating a desire not to rely on negotiations between Brussels and Anthropic.

A three-year agreement was signed between BNP Paribas and Mistral AI in February, covering all models developed by Mistral. Mistral's argument is based on the creation of a sovereign model, which, even if less powerful than Mythos, would provide valuable protection and enhance European autonomy.