Overview
- BNP Paribas announced on Tuesday that it expanded its multi‑year collaboration with French startup Mistral AI to prioritise cybersecurity after industry concern over advanced models that can discover many software flaws quickly.
- Under the expanded deal Mistral will develop a dedicated cyber model for European banks and embed engineering teams inside BNP to co‑develop benchmarks, tooling and joint research for regulated use cases.
- BNP already runs Mistral models for internal tools, virtual assistants and compliance and plans a generative AI employee assistant in 2026 while keeping key systems on‑premises to protect customer data and meet EU rules.
- The move responds to warnings about models such as Anthropic’s Mythos that can scan code and suggest exploits far faster than human teams, a reported test result that one outlet put at a 72% exploit success rate for Mythos.
- The partnership reflects a wider European shift toward regionally based AI providers that ease data‑sovereignty and regulatory risks and could prompt faster, machine‑speed vulnerability remediation across banks.