AI in action · Regulation
A Former FDIC Chair Told China's Bank Regulator That AI Must Come With "Governance, Accountability and Risk Management"
Sheila Bair, former chair of the US FDIC, told the annual meeting of the international advisory council of China's National Financial Regulatory Administration that any use of AI in finance must come with clear governance, accountability and risk management, China Daily reported on October 23, 2026. Fellow member Andrew Sheng warned of "hallucinations" and asked for real-time supervisory tools.
China's banking and insurance regulator, the National Financial Regulatory Administration, keeps an International Advisory Council of former officials and financiers from outside the country, and its 2026 meeting in Beijing, held on Monday and Tuesday of last week alongside the Financial Street Forum, spent much of its time on artificial intelligence. China Daily's Jiang Xueqing reported the discussion on October 23. The most direct statement came from Sheila Bair, who chaired the US Federal Deposit Insurance Corporation during the 2008 financial crisis: "It is absolutely essential that any use of AI be accompanied with clear governance, accountability and risk management."
Bair's list of risks was the one supervisors on both sides of the Pacific now recite, which is part of its interest. AI helps banks defend themselves against cyberattacks but gives attackers new capabilities too. Models can be opaque, can carry bias into credit decisions and can be used for fraud. And the industry's reliance on a few AI providers creates concentration risk, the same concern the Bank of England, the ECB and the US agencies have raised. Her criticism was of behaviour rather than technology: firms around the world are rushing to adopt AI, she said, and often put risk management behind commercial gains. Her recommendations were that financial institutions adopt AI cautiously until regulators understand its risks and how to contain them, and that the industry build information-sharing protocols to spot and counter malicious uses.
Andrew Sheng, a council member and adviser to Hong Kong's chief executive, focused on the regulators' side of the problem. Poor data and weak governance, he said, can produce flawed models whose outputs, including "hallucinations," lead to misleading analyses or signals, and supervisors who do not understand large language models, distributed ledgers or privacy-enhancing computation risk regulatory failure. He asked for real-time supervisory tools that can detect algorithmic opacity, market anomalies and hidden connections, clearer rules on data ownership and access, and coordinated cross-border crisis protocols, since divergent AI standards, worsened by geopolitical tensions, can let risk travel where rules do not.
The meeting's own conclusions were more general. According to China Daily, the council agreed to strengthen policy coordination and cross-border supervisory cooperation, to encourage institutions to manage emerging risks while using new technologies including AI to improve efficiency, to improve early-warning and response mechanisms, and to give priority to preventing systemic risk. None of this is a rule, and the NFRA did not announce one.
What makes the exchange worth reading is the setting. A former head of the US deposit insurer giving the same advice to Beijing's regulator that US, European and Japanese supervisors are giving their own banks this month is a sign that the AI risk agenda, at least, has become common ground. For investors in Chinese banks and insurers, the practical signal is that Chinese supervisors are hearing, from people they invited, that model governance and concentration on a few AI vendors are things to examine before the next crisis rather than after it.
Sources: China Daily (Jiang Xueqing) — Financial regulators urged to strengthen AI oversight (Oct 23, 2026)