Mistral AI in talks with European banks to develop Mythos as a replacement for internet security models

Bloomberg, citing people familiar with the matter on May 14, reported that French AI startup Mistral AI is developing an autonomous in-house cybersecurity AI model. The company has already held discussions with European banking institutions to offer alternative options for financial institutions that cannot access Anthropic PBC’s Mythos. Mistral worked with bank customers to use AI to identify security vulnerabilities before Mythos was released.

Current status of confirmed access restrictions for Mythos

According to Bloomberg’s report, Anthropic Mythos currently uses a restricted access model, granting access only to a small number of partners, including some technology companies, banks, and cybersecurity firms. These partners are currently testing its capabilities and strengthening their own defenses. Early analysis suggests that Mythos has the ability to carry out autonomous attacks. Anthropic representatives did not respond to Bloomberg’s request for comment. Mistral representatives also declined to comment.

Mistral cybersecurity model: confirmed status and CEO remarks

Arthur Mensch, CEO of Mistral AI, said during a hearing at the French National Assembly: “We must have control over this technology. You can’t let Mythos scan the source code of the French armed forces. This creates an irreparable dependency; we must find a solution.” Mensch said that Mistral and other U.S. and Chinese companies’ technologies already have the ability to detect cybersecurity vulnerabilities, and that discussions around Mythos would be characterized as “spreading panic.”

According to people familiar with the matter cited by Bloomberg, the release date for Mistral’s cybersecurity AI model has not been determined, and discussions with European banking institutions are still ongoing.

OpenAI GPT-5.5-Cyber: confirmed available to European institutions

According to Bloomberg, OpenAI has launched a cybersecurity-focused model, GPT-5.5-Cyber, and has already provided access to several large European companies, including Spain’s bank BBVA SA.

Mistral AI’s financial background and current financial clients

Mistral AI achieved a $12 billion valuation in September 2025 after raising €1.3 billion in funding led by Dutch chipmaker ASML Holding NV. Bloomberg’s report confirmed that its existing financial clients include HSBC and BNP Paribas.

FAQ

Why does Anthropic Mythos restrict access, and which institutions does it currently allow?

According to Bloomberg’s report, early analysis indicates that Mythos can discover and launch autonomous cyberattacks at an unprecedented speed and scale, prompting Anthropic to restrict access after its release. At present, it is only open to a limited number of partners, including some technology companies, banks, and cybersecurity firms, for functionality testing and defense hardening. Anthropic representatives did not respond to the request for comment.

When will Mistral’s cybersecurity AI model be officially released?

According to people familiar with the matter cited by Bloomberg, the release date for Mistral’s cybersecurity AI model has not been confirmed. The people familiar with the matter said the model is still under development, and discussions with European banking institutions are ongoing as well, but there is no publicly available specific launch timeline.

What confirmed AI cybersecurity options are European banks currently able to use?

As of Bloomberg’s May 14, 2026 report, the confirmed options for European banks include: OpenAI GPT-5.5-Cyber (already provided access to European companies such as BBVA SA); Anthropic Mythos (only open for testing to a small number of partners). Mistral’s commercial version has not been released yet and remains in the collaborative development stage with customers.