BNP Paribas SA is coordinating with Mistral AI and other partners to address cybersecurity challenges arising from a new generation of artificial intelligence models, including Anthropic PBC's Mythos.
At a press conference in Paris on Tuesday, BNP's Chief Information Officer Marc Camus emphasized that the response should not focus only on access to Mythos, noting that comparable models exist at other firms. His remarks framed the bank's approach as covering a broader field of emerging AI threats rather than a single product.
Mistral has held talks with European banks about deploying a dedicated response to Mythos, according to reporting on the discussions. That limited-access AI model - released by Anthropic in April - has prompted concern among financial institutions because Anthropic warned it represents a watershed moment for security due to its capability to detect software security flaws. Mythos has not been made available to the general public; Anthropic and a select set of mostly US partners are currently testing its capabilities.
Separately, Mistral has been developing its own cybersecurity-focused AI model. Corentin Petit, Mistral's head of solutions, said on Tuesday that the company is engaging directly with finance clients on specific cybersecurity use cases, although he noted that no formal announcement has yet been made. Those engagements with banks appear to include discussions about how to deploy Mistral's tools within financial institutions.
The relationship between BNP and Mistral was recently renewed and extended under a three-year agreement. The scope of that collaboration reaches beyond cybersecurity. BNP Paribas plans to apply AI to automate and replace certain tasks, with cited examples including virtual assistants for retail customers and AI-driven scenario planning tools for bankers. The bank also stated it is not restricted to working exclusively with Mistral, but it did not disclose the names of any other technology providers involved.
Implications for markets and sectors
- Financial services - banks face new operational and security considerations as advanced AI models surface potential software vulnerabilities.
- Cybersecurity and AI vendors - demand for models and services focused on threat detection and mitigation may rise as institutions evaluate responses to limited-access models like Mythos.
Context and limitations
The information available notes ongoing discussions and product development but does not indicate public availability of Mistral's cybersecurity model or list other partners by name. BNP's statements highlight strategic intent and contractual extension but stop short of detailed deployment timelines or technical specifications.