The Bank of England is discussing whether to include artificial intelligence as a central risk in its next system-wide exploratory scenario, according to sources familiar with the matter. The current exercise focuses on private markets and is set to conclude next year, but AI is rising on the regulatory agenda amid concerns over valuation bubbles, cyber risks, and the potential for AI-driven trading models to amplify market shocks.
Officials are concerned that AI agents used in trading could prompt herding behaviour, increasing volatility if models respond similarly to market events. There is also regulatory attention on the risk of AI exposing weaknesses in cyber defences and the possibility of an AI stock valuation bubble bursting with wider economic effects. The Bank's current scenario already includes some AI disruption, but officials say the scope of concern has widened since the exercise began.
Governor Andrew Bailey has said AI could boost growth but has also warned of its risks, particularly in cyber security. The Bank is still considering its options for the next stress test, with no final decision yet on whether AI will be the main focus.
Senior executives in banking and insurance should prepare for more scrutiny of AI risk management as regulators assess how technology could affect financial stability.
Cybercriminals are shifting from traditional ransomware and data breaches to more sophisticated fraud and business email compromise (BEC), according to executives at the NetDiligence Cyber Risk Summit. Christine Mapes of Travelers said that about 50% of the insurer’s cyber claims now involve financial fraud or BEC, with a 57% increase in such claims from the first half of 2025 to the first half of 2026. Claim severity has also risen by about 30% since 2023.
Tactics have evolved: attackers now use IT and bank fraud manipulation, often convincing victims to hand over access voluntarily. Artificial intelligence is fuelling this trend, enabling criminals to bypass language barriers and improve impersonation, according to Integreon’s Blake Feldman and Westfield Specialty’s Sara Trokan. Fraudsters increasingly combine email, audio, video, and social media to create convincing scams, sometimes tricking staff into making large payments.
While AI is also used to strengthen cyber defences, the panel agreed that basic verification processes remain effective. Lindsay Nickle of Constangy suggested that raising awareness and using low-tech checks, as was done with social engineering threats in the past, is still necessary.
Insurers and regulated firms should review their fraud controls and staff training as AI changes the threat profile. Monitoring new fraud trends and reinforcing verification procedures will be important as attackers adapt.
