TL;DR:

• The Bank of England has confirmed it will investigate the impact of AI agents demonstrating correlated or “herding” behaviour in financial markets, a direct response to the Treasury Committee.
• The FCA will share AI best-practice examples with financial services firms after MPs criticised the clarity of current guidance.
• HM Treasury has declined to commit to a 2026 deadline for bringing major AI and cloud providers into the Critical Third Parties Regime, drawing public rebuke from committee chair Dame Meg Hillier.

The Treasury Committee today published formal responses from the Bank of England, HM Treasury and the Financial Conduct Authority to its report on AI in financial services. The Bank’s agreement to run AI-specific stress testing is the headline concession, aimed at answering MPs’ concerns that synchronised AI agent behaviour could amplify market stress.

What each regulator has committed to

The Bank will monitor the Financial Policy Committee’s view on HM Treasury’s use of the Critical Third Parties Regime — the framework that can bring cloud and AI providers under direct financial regulatory oversight. The FCA will publish practice examples to help firms align AI deployment with existing conduct rules, a concrete response to industry complaints that rule application to AI workflows remains ambiguous. HM Treasury, by contrast, gave no timeline for extending the CTP Regime to cover AI and cloud providers before the end of 2026.

The Mythos accelerant

Dame Meg Hillier cited Anthropic’s Project Mythos in her statement, noting how quickly the risk landscape is moving. The same week ministers at the IMF spring meetings debated Mythos’s cybersecurity implications, and UK banks were preparing controlled access under FCA, HM Treasury and National Cyber Security Centre coordination. Hillier described herself as “pleased” with the Bank’s response but “perplexed” at Treasury inertia — an unusually direct criticism of a department that still holds the lever for designating AI providers as critical third parties.

What it means for UK financial services

For UK banks and insurers, the immediate signal is that AI-agent deployment in trading and asset management now sits under explicit supervisory scrutiny. Firms building or procuring agentic trading workflows should expect to be asked how correlated behaviour across their own estate — and across the industry — has been modelled. For third-party AI providers, the CTP delay is a reprieve on paper only; the Bank’s monitoring role and the committee’s watchful posture mean the designation question is postponed, not settled.

Looking forward

The next inflection point is the FCA’s practice examples, which will reveal how prescriptively the regulator wants to shape AI deployment. If the guidance stays principles-based, compliance burdens fall on individual firms to defend. If it becomes more specific, expect pressure for a similar lead from the Prudential Regulation Authority on operational resilience.