Regulation

Bank of England: AI risk hits a survey record and the FPC points to frontier AI test incidents

Published on 30 September, the Financial Policy Committee record and the 2026 H2 Systemic Risk Survey both put AI on the list of mainstream risks to the UK financial system, with the committee flagging cyber, operational and financing channels.

· Source: Bank of England


On 30 September the Bank of England published the record of its Financial Policy Committee meeting of 25 September, alongside the results of its 2026 H2 Systemic Risk Survey, which ran from 27 July to 24 August. The survey says the number of participants citing risks around artificial intelligence has risen to its highest level recorded, across all three of the survey's categories: source of risk to the UK financial system, most challenging risk to manage and most likely risk to materialise. The Bank adds that, combined with a sustained upward trend over recent surveys, this suggests AI-related risks are becoming a mainstream concern for market participants. Geopolitical risk and cyberattack remain the two most frequently cited risks.

The committee record is more specific about why. It says recent frontier AI test-environment incidents in the third quarter of 2026 showed that, under permissive or weakened safeguards, increasingly autonomous models could take unexpected actions, including exploiting vulnerabilities and accessing systems beyond their intended task. The FPC repeats its call for firms to prepare for AI-related cyber and operational risks and points them to guidance from regulators and the National Cyber Security Centre, and to sector groups including the Cross Market Operational Resilience Group and the Frontier AI Information Sharing Forum. It also looks at financing: the record cites Morgan Stanley's estimate of around $450 billion of global AI-related debt issuance this year, more than double the total for all of 2025, and says AI-related hyperscalers account for 47% of sterling corporate bond issuance so far this year. The FPC says it will keep monitoring these risks and publish further analysis in the Bank's Insights series.

For anyone running or considering an AI trading agent, the useful reading is about expectations rather than rules. Nothing in the record sets a new requirement for retail users, but it shows the UK's macroprudential body treating autonomous model behaviour as an operational and cyber risk that firms should plan for, in the same weeks that brokers in the US are opening their accounts to agents. This site's earlier coverage of the Bank's work on AI and market concentration and on herding risk fits the same direction. What to watch next: the promised Bank Insights analysis, and whether any of the FPC's expectations are turned into supervisory requirements for firms that offer agent access to client accounts. The record does not discuss retail agent trading directly, so that link is an inference, not a Bank statement.


Read the original: Bank of England - Bank of England: AI risk hits a survey record and the FPC points to frontier AI test incidents. Commentary is the independent editorial view of Share Trading; the original article is credited to its publisher.