The Bank of England has put AI-related debt and circular financing squarely on the financial stability agenda, showing that AI’s impact on finance now covers systemic risk as well as productivity. Alongside that, banks are shifting towards on-premises and governed AI infrastructure, and a new market for AI-liability insurance is emerging out of Lloyd’s. Regulation remains in flux, with EU high-risk AI obligations pushed back to December 2027.
Top story: The Bank of England’s Financial Policy Committee warned that rapid, circular AI-related debt financing could amplify losses and threaten UK financial stability.
Bank of England warns AI debt boom threatens financial stability
FStech · Risk
In the record of its 25 September meeting, the Bank of England’s Financial Policy Committee flagged the growing indebtedness of AI firms and their opaque, ‘circular’ financing arrangements as a risk that could amplify losses. It noted that sterling-denominated AI debt accounted for 47% of GBP corporate bond issuance this year. For risk and investment professionals, AI exposure now sits in credit and bond portfolios, not just in tech equities.
Go.AI raises $85M as banks seek on-premises AI control
FinAi News · Finance
On-premises AI provider Go.AI raised $85 million in a Series A led by Updata Partners to build a large language model designed for financial services workflows. Its CEO said more financial institutions are bringing AI in-house to avoid dependence on a single cloud or model provider, with interest rising after a simultaneous outage among major LLM providers on 3 September. The story points to a growing resilience and sovereignty concern in bank AI strategy.
https://finainews.com/lending/fintech-funding-go-ai-raises-85m/
EU delays high-risk AI Act rules for finance to December 2027
Legalithm · Regulation
According to this compliance guide, the Digital Omnibus deferred full enforcement of high-risk AI obligations from 2 August 2026 to 2 December 2027. Credit scoring, life and health insurance pricing and AML profiling are explicitly classed as high-risk under Annex III. Firms get more time, but other guides still cite the August 2026 date, so practitioners should confirm the current timeline with legal counsel.
https://www.legalithm.com/en/blog/ai-act-financial-services-banking-insurance-compliance
Bank of England to stress-test AI agents for market ‘herding’
UK Parliament Treasury Committee · Regulation
Responding to MPs, the Bank of England confirmed plans to investigate how AI agents in trading markets might behave in correlated ways, known as ‘herding’. HM Treasury, however, did not commit to bringing major AI and cloud providers into the Critical Third Parties regime before the end of 2026, and the FCA agreed to share best-practice examples. This is from April 2026 but remains the key UK policy backdrop for firms deploying agentic AI in trading.
Lloyd’s market writes cover for AI hallucinations and agent errors
AgentMarketCap · Strategy
By April 2026, Lloyd’s syndicates working with Armilla and AIUC, plus Munich Re, were writing affirmative policies that pay out when AI agents fabricate outputs, leak data or breach regulation. Armilla’s policy reportedly covers up to $25 million per organisation, including fines under the EU AI Act. For insurers and corporate risk managers, it shows AI liability becoming a distinct, priced line of business. This is an older, secondary source, so treat details with some caution.