France-based Mistral AI is in discussions with European banks about a cybersecurity-focused AI model it has been developing, according to sources cited by Bloomberg. The talks come as European financial institutions remain largely locked out of access to Mythos, the cybersecurity-capable AI model developed by Anthropic that has raised widespread concern about AI's capacity to identify and exploit software vulnerabilities at scale.
Mistral has been working with banking clients on AI-assisted vulnerability detection before Mythos's release, but is now developing a more broadly deployable version of the product, according to people familiar with the matter who asked not to be identified. The company declined to comment. Additionally, a timeline for the model's release has not been disclosed.
The Mythos context
Anthropic's decision to limit access to Mythos has prompted a global response among AI developers and regulators. Early analysis by the UK's AI Safety Institute found the model capable of autonomous attacks. Access to Mythos has been restricted to a select group of partners, including technology firms, banks, and cybersecurity companies currently testing its capabilities. European institutions have largely been excluded from this group, a situation that has prompted concern about a transatlantic security gap and accelerated efforts to find alternative solutions.
The White House has separately been considering an executive order to require safety vetting of new AI models ahead of public release. OpenAI has also entered the cybersecurity AI space with GPT-5.5-Cyber and has granted access to several large European firms, including Spain-based BBVA.
Mistral's positioning
Mistral is expected to position its model as a sovereign, lower-risk alternative for European institutions that are unable or unwilling to rely on US-developed tools for sensitive security work. The company counts HSBC and BNP Paribas among its financial sector clients, providing an existing distribution base for a cybersecurity product targeting European banks.
Market and regulatory context
The emergence of AI models capable of identifying cybersecurity vulnerabilities at speed has created a dual challenge for financial institutions: the need to access such tools defensively to shore up their own systems, while managing the risk that the same capabilities could be used offensively against them. For European banks, the restricted availability of Mythos has made that defensive use case harder to address, creating demand for a European alternative that can be deployed within local data sovereignty and regulatory frameworks.
Mistral was valued at nearly EUR 12 billion in September 2025 following a EUR 1.7 billion investment round led by ASML. The development of a dedicated cybersecurity model would represent a significant expansion of its financial services offering at a moment when AI-driven cyber risk has become a priority concern for regulators across Europe, including Germany's BaFin, which recently announced a new division for targeted IT inspections at financial firms.
