Key facts
- EU banks must submit AI risk management plans by the end of October.
- Plans should detail strengthening internal systems and vetting external technology providers.
- Advanced AI models pose risks to the confidentiality, integrity, and resilience of financial systems.
- The European Systemic Risk Board warned that frontier AI models should be treated as a systemic risk.
- The European Commission will also unveil an action plan on AI risks.
The European Central Bank's top banking supervisor, Claudia Buch, has directed the European Union's largest lenders to present comprehensive plans by the end of October detailing how they will address the risks posed by advanced artificial intelligence models. Buch emphasized the need for banks to bolster their internal systems and ensure the suitability of external technology providers.
These instructions come in response to growing concerns that sophisticated AI, such as Anthropic's Mythos, could exploit software vulnerabilities at unprecedented speeds, potentially destabilizing the financial system. Buch highlighted that such AI capabilities could significantly impact the confidentiality, integrity, and resilience of banks' information technology infrastructure.
While no explicit sanctions are planned for non-compliance, the ECB intends to use these submitted plans to benchmark banks' approaches to AI risk management and engage in follow-up discussions. The directive also urges bank CEOs to modernize their infrastructure and enhance crisis management and recovery mechanisms.
Concurrently, the European Systemic Risk Board (ESRB) issued a warning, classifying frontier AI models as a potential source of systemic risk for the finance industry. The ESRB cautioned that cyber incidents originating from these models, if spread through critical operational systems like payment and settlement, could severely disrupt the financial system and evolve into structural vulnerabilities without coordinated EU action.
The European Commission is also expected to release its own action plan addressing AI risks, signaling a broader bloc-wide effort to ensure the safety and testing of advanced AI models.