Sam Woods, chief executive of the Prudential Regulation Authority (PRA), told a UK Finance growth conference that the latest artificial intelligence models — including Anthropic Mythosa and ChatGPT 5.5 Instant — could cause "very significant disruption" to the financial sector, according to Reuters.
Woods explained that these AI models are increasingly effective at uncovering vulnerabilities in financial systems. That capability forces banks to carry out necessary fixes more quickly; however, the PRA chief said that the resulting haste is currently a principal cause of operational incidents in the financial system.
He urged financial institutions to strengthen basic cyber‑security procedures and to speed up their incident response times. Woods also argued that AI‑based defensive solutions will become increasingly important to enable faster and more effective responses.
Cyber‑security experts cited in the report warned that the new technology poses particular challenges for banks with outdated IT systems. Anthropic released its Mythos model in April, and it is presently available only to a limited number of business customers.
At the same time, a cyber‑security working group co‑chaired by the Bank of England concluded last month that the sector is prepared for these challenges. Together, the PRA’s comments and the working group’s finding suggest that while the risks from advanced AI are real and material, both regulators and the industry are actively addressing them.
Why this matters
- Rapidly evolving generative AI models can reveal system‑level weaknesses that require urgent remediation.
- The rush to fix vulnerabilities can itself increase the risk of operational disruption.
- Improving basic cyber practices, reducing response times, and deploying AI‑driven defenses are likely to be key to preserving banks’ operational resilience.
The measures Woods called for, and the PRA‑led working group’s assessment, together frame the current regulatory and industry response to the new classes of risk introduced by advanced AI.
