Bank of Spain calls for access to advanced AI tools, flags cyber risks

• New AI tools like Anthropic's Mythos model could sharply reduce the time between vulnerability detection and malicious exploitation, posing significant challenges to the banking industry
• Under adverse scenarios, such AI models could have negative systemic impact by enabling synchronized cyberattacks across the global financial sector and broader economy
• The Bank of Spain emphasized the need for closer collaboration between regulators, financial institutions, and technology providers to increase resilience and prevent attacks

Summary: Bank of Spain Warns of AI-Driven Cybersecurity Risks

The Bank of Spain issued urgent calls for enhanced international cooperation and broader access to protective AI systems in its semiannual financial stability report released May 14. The central bank specifically highlighted concerns about Anthropic's AI models—the vulnerability-detection tool Mythos and its protective counterpart Glasswing.

Key Concerns:

The report warns that advanced AI tools like Mythos could significantly reduce the time between vulnerability detection and malicious exploitation, making cyberattacks easier to execute. Under adverse scenarios, these models could enable synchronized cyberattacks across the global financial sector, creating systemic risk.

Recommendations:

The Bank of Spain emphasized the need for:

• Stronger international regulatory coordination
• Wider access to protective AI models like Glasswing (currently limited-access)
• Enhanced collaboration between regulators, financial institutions, and technology providers

Market Implications:

Financial Stability Director Daniel Pérez Cid confirmed discussions with Spanish lenders regarding AI-related risks. The banking sector's cyber resilience will increasingly depend on rapid vulnerability response and the ability to anticipate emerging threats. The report suggests that restricted access to protective AI tools while offensive capabilities proliferate could leave financial institutions vulnerable.

Broader Context:

This warning comes as global regulators grapple with AI governance. Parallel developments include U.S.-China discussions on AI guardrails scheduled for their Beijing summit, focusing on protocols to prevent non-state actors from accessing the most powerful AI models.

The advisory underscores growing regulatory concern that AI advancement is outpacing defensive capabilities in the financial sector.