This week’s dominant theme is accountability: who audits the AI systems now embedded inside banks, insurers, and hedge funds, and what happens when the answer is unclear. The EU AI Act’s transparency provisions became enforceable on 2 August, the UK PRA’s updated model risk rules are live, and a landmark Hazeltree report shows hedge funds are quietly repricing their AI crowding risk — all in the same fortnight.
Top story: Global Banking & Finance reveals a structural governance blind spot: banks are deploying AI agents they cannot fully audit, observe, or control — and regulators are starting to ask hard questions.
Banks Running AI They Cannot Fully Audit, Regulators Warn
Global Banking & Finance Review · Risk
A major investigation published this week finds that AI is no longer just an analytical layer inside banks — it has become operational infrastructure making real-time decisions on credit, fraud, and liquidity. The UK PRA’s SS1/23 model risk rules, whose updated version came into full effect in April 2026, explicitly require banks to identify and manage AI and ML model risks, while the European Banking Authority’s June 2026 risk assessment warned that AI simultaneously improves detection and introduces new failure modes. The core problem: traditional model risk management assumes you can examine what a model does before it acts — a premise that agentic AI systems fundamentally break.
EU AI Act Transparency Rules Now Live: Chatbots Must Disclose AI Status
Lanware · Regulation
From 2 August 2026, a new tranche of EU AI Act obligations became enforceable — including transparency rules requiring bank and insurer chatbots to disclose they are AI systems, and regulators gaining full penalty powers. Crucially, the European Commission’s Digital Omnibus package has pushed the high-risk obligations for Annex III systems — including credit scoring and insurance pricing — back to December 2027, giving firms breathing room but creating a confusing two-tier enforcement timeline. UK financial services firms with no EU footprint may not be directly in scope, but any firm serving EU customers or running AI outputs in EU markets falls within reach.
Hedge Funds Quietly Rotate Out of Crowded AI Trades
Hedgeweek · Finance
New data from Hazeltree’s Crowding Report — covering anonymised securities-finance data from more than 700 hedge funds across approximately 16,000 securities — shows global funds rotated within the AI trade during July, cutting exposure to the most crowded names and moving toward companies with clearer evidence that AI investment is translating into revenue and earnings. Analysts describe the shift as a refinement of risk rather than a broad retreat, but it follows a brutal month in which leveraged AI positions unwound sharply. For institutional risk managers, the report is a live stress test of what happens when too many funds crowd into the same thesis with borrowed money.
French Insurers Push AI From Pilots to Live Underwriting Decisions
FinTech Global · Strategy
At Earnix’s first France-focused Eˣcelerate event in Paris, executives from Crédit Agricole Assurances and Alptis discussed where automated decisions, technology-assisted judgement, and human expertise should sit within underwriting workflows — a conversation that signals French insurers are moving beyond experimentation into production deployment. The central tension is transparency: how to introduce greater automation without losing human oversight in decisions that directly affect pricing and policy issuance. The discussion is directly relevant to UK and European brokers, as a separate Zywave survey of over 1,400 US employers found that ‘failure to leverage modern technology and AI-powered tools’ has for the first time entered the top reasons employers would switch their insurance broker.
https://fintech.global/2026/08/10/french-insurers-move-ai-from-pilots-to-the-front-line/
Fintech Funding Surges 23% as Investors Concentrate Bets on AI Infrastructure
Crunchbase News · Finance
Fintech funding rose 23% in H1 2026 as investors concentrated capital on AI-native companies compressing underwriting, fraud detection, and advisory workflows that previously required teams of analysts working for weeks. Traditional financial institutions are finally bringing AI into core operations, but analysts caution the shift is opening as much risk as opportunity — particularly around cybersecurity and the governance layer, which one senior investor described as now being ‘just as important as the AI itself.’ The IPO market for fintechs has been notably quieter in 2026 than in 2025, with only three foreign companies listing in New York in H1, suggesting capital is staying private and concentrating in fewer, higher-conviction bets.
https://news.crunchbase.com/fintech/funding-rises-deals-slump-h1-2026/