歐洲央行警告 AI 恐引發金融危機並呼籲全球監管

Generative AI can sharply increase the scale and precision of phishing emails, malware and identity-impersonation attacks, exposing banks’ existing cybersecurity systems to new risks. The European Central Bank has therefore instructed banks under its direct supervision to strengthen their defenses, seeking to prevent a breach at one institution from disrupting payments, deposits and financial stability across the eurozone.

The ECB has required all banks under its direct supervision to submit formal remediation plans by October 31, detailing how they will close gaps in their defenses against AI-driven cyberattacks. The requirement also covers eurozone subsidiaries of US banks including JPMorgan Chase and Goldman Sachs, with supervisors focusing on governance, detection and response capabilities.

Agentic AI can independently analyze information, make decisions and execute trades with minimal human intervention. If multiple financial institutions use similar models, erroneous signals could trigger synchronized buying or selling, amplifying price swings and liquidity gaps when markets are under stress. The Bank of England, European Central Bank and UK Financial Conduct Authority therefore view the technology as a financial-stability issue.

On June 30, 2026, Bank of England Deputy Governor Sarah Breeden told the ECB’s annual forum in Sintra, Portugal, that authorities should consider market-wide circuit breakers or emergency stop mechanisms. ECB President Christine Lagarde warned on July 2 that risks were intensifying, while FCA Chief Executive Nikhil Rathi said traditional regulatory cycles could not keep pace with AI systems evolving over weeks or months. The Bank for International Settlements had also warned on June 28 that a sharp fall in AI asset prices could trigger financial contagion.

The European Central Bank is reviewing banks’ credit exposure to AI-related industries, including data centers and AI developers. AI infrastructure requires substantial capital, while concentrated lending, valuation volatility and rapid technological obsolescence could transmit industry risks to banks’ asset quality. The sector has therefore become a focus for financial supervisors.

The latest survey examines the scale of banks’ lending to the AI sector, borrower risks and internal controls. The ECB has also held a workshop to understand how banks use artificial intelligence. Reports did not disclose when the survey began, how many banks were involved, the total credit exposure or a completion timetable, indicating that the review remains at the risk-mapping and information-gathering stage.