This week’s stories reveal a global finance sector moving decisively from AI experimentation to regulated deployment, with regulators in the UK, US, and EU all tightening governance frameworks simultaneously. The twin pressures of an imminent EU AI Act compliance deadline (August 2026) and rising AI-enabled financial crime are forcing institutions to treat explainability and auditability as non-negotiable. Meanwhile, agentic AI is quietly rewriting operational benchmarks — from Lloyd’s underwriting turnaround times to private equity deal sourcing — creating a widening gap between early movers and the rest of the market.
Top story: The UK FCA’s Mills Review formally identifies managing AI autonomy as financial regulation’s central challenge, signalling a structural shift in how the sector will be supervised.
FCA’s Mills Review: Managing AI Autonomy Is Finance’s Defining Regulatory Challenge
Banking Dive · Regulation
The UK Financial Conduct Authority’s Mills Review concludes that financial services is shifting from human-led to AI-enabled, with agentic AI now deployed at 62% of financial services firms — 93% of which grant those agents autonomy. The FCA has committed to building an AI-enabled agentic supervisory model of its own and to adapting its regulatory framework to manage the transition, explicitly naming ‘striking the balance between enabling delegation and managing autonomy’ as its central challenge. For practitioners, this signals that the era of light-touch AI oversight in the UK is ending, and governance documentation will face much closer scrutiny.
https://www.bankingdive.com/news/ai-reshape-financial-services-regulation/824563/
EU AI Act Compliance Clock Ticking for Banks and Insurers: August 2026 Deadline Looms
Global Regulation Tomorrow · Regulation
With the EU AI Act’s high-risk compliance deadline for financial institutions falling on 2 August 2026, European regulators are now incorporating AI Act checks directly into existing supervisory review processes (SREP) — meaning AI governance will be assessed alongside capital adequacy and operational resilience. AI systems used for credit scoring, insurance underwriting, AML, and fraud detection are all classified as high-risk, exposing non-compliant institutions to fines of up to €30 million. The European Commission’s July 2025 guidance clarified that simpler rule-based and statistical models fall outside scope, providing some relief, but institutions with adaptive or inference-based models must act now.
AI-Enabled Financial Crime Hits $579bn as Arms Race Goes Bidirectional
Hedge Think · Risk
Global financial crime losses reached $579.4 billion in 2025, with 67% of banks and fintechs reporting rising fraud rates — and the same AI capabilities powering fraud detection now being weaponised by criminals through deepfake voice attacks, synthetic identities, and adversarial inputs designed to evade ML classifiers. A Hawk/Chartis survey of 250 banks and fintechs found that while 90% of banks now encourage AI use in financial crime compliance, only 16% have moved beyond pilots to full operational deployment. The practical implication: firms with immature AI fraud stacks face asymmetric risk as attacker sophistication outpaces defensive tooling.
Hiscox Cuts Lloyd’s Underwriting Turnaround from Three Days to Three Minutes
Medium / LMA Research · Generative AI
A detailed analysis of Lloyd’s of London market adoption reveals that Hiscox deployed a live agentic AI system for its Sabotage & Terrorism line in August 2024, reducing quote turnaround time from three days to three minutes — a 99% reduction — without adding headcount. Despite this, a Lloyd’s Market Association survey of 81 firms found only 14% have deployed agentic or generative AI in underwriting, with the remaining 86% still watching or experimenting. The divergence signals a rapidly widening competitive gap inside the London specialty market, with early movers compounding their advantage at speed.
PE Firms Pour Capital Into AI but Only 20% Have Live Production Results
Ropes & Gray / EY · Strategy
A confluence of industry data shows private equity’s AI adoption has reached a critical inflection: 86% of dealmakers now use generative AI in workflows, 84% of PE firms have appointed a Chief AI Officer, and 88% have invested over $1 million in generative AI. Yet Bain’s 2025 Global PE Report finds only around 20% of portfolio companies have operationalised a generative AI use case with concrete results, exposing the gap between licence spend and real workflow integration. Several major firms — including General Atlantic and Schroders Capital — have established dedicated AI roles on investment committees, pointing to where genuine competitive differentiation is beginning to emerge.
https://www.ropesgray.com/en/insights/alerts/2025/08/artificial-intelligence-h1-2025-global-report