The dominant theme this week is the collision of agentic AI deployment with a tightening regulatory vice: the EU AI Act’s August 2026 high-risk deadline is weeks away, the UK FCA is openly rethinking its oversight model, and banks are being asked to prove — not just claim — that their AI systems are explainable and accountable. Simultaneously, real commercial momentum is building, with Anthropic launching a dedicated financial-services agent marketplace, a Goldman Sachs-backed fintech raising $110M for AI-driven underwriting, and insurers being challenged to break out of perpetual pilot mode.
Top story: The UK’s FCA chief publicly declared that traditional rule-making ‘will not work everywhere’ for AI — signalling a fundamental shift in how Britain’s most powerful financial regulator will govern AI in banking and investment.
FCA Chief Rethinks Regulation as Agentic AI Hits Financial Services
Digital Watch Observatory · Regulation
FCA CEO Nikhil Rathi told a major techUK conference that over 80% of financial services firms are already using or adopting AI, shifting policy focus from adoption to large-scale deployment — and that existing legislative frameworks alone cannot keep pace. The FCA is exploring using agentic AI itself as a ‘first responder’ for wholesale market abuse monitoring, and plans to publish the Mills Review on AI in retail financial services imminently, alongside formal guidance on good and poor AI practice. For compliance and risk teams, this signals that personal accountability under SM&CR will increasingly extend to automated decisions.
https://dig.watch/updates/fca-ai-regulation-financial-services
EU AI Act High-Risk Deadline Hits Banks and Insurers in August
Alice Labs · Regulation
With the EU AI Act’s full compliance deadline for high-risk AI systems falling on 2 August 2026, banks and insurers using credit scoring, AML transaction monitoring, and insurance underwriting AI must now have documented conformity assessments in place or face fines of up to €30 million. The European Banking Authority has confirmed the majority of AI use cases at supervised institutions fall into the high-risk category — meaning this is not a niche compliance edge case but the mainstream of financial AI. Institutions are also reminded they cannot outsource compliance to AI vendors, retaining full responsibility for how systems are used in practice.
https://alicelabs.ai/en/insights/eu-ai-act-for-financial-services
Anthropic Launches Finance-Specific Agent Marketplace for Banks and Insurers
Anthropic · Generative AI
Anthropic has launched a dedicated financial services marketplace offering pre-built agent templates for banks, asset managers, hedge funds, and insurers, integrating data from Moody’s, Morningstar, PitchBook, Verisk, and SS&C Intralinks directly into Claude workflows. The platform covers front, middle, and back office tasks — from credit analysis and underwriting to deal-room diligence and market abuse monitoring — with Claude Opus 4.7 leading the industry on the Vals AI Finance Agent benchmark. This marks a significant step-change from general-purpose AI tools to purpose-built, data-connected agents for regulated financial workflows.
Goldman-Backed Taktile Raises $110M to Automate Underwriting and KYC
Crescendo AI / AI Funding Tracker · Finance
Taktile, whose customers include Monzo, Allianz, and Rakuten Bank, closed a $110M Series C led by Goldman Sachs Growth Equity alongside Balderton Capital and Index Ventures, bringing total funding to $184M. The platform automates loan approvals, fraud detection, insurance claims, and KYC processes, claiming 95% automation in business underwriting and a 75% reduction in AML false alarms. The raise is a strong signal that investors see AI-native decisioning infrastructure — rather than point AI tools — as the foundational layer for the next generation of financial services operations.
https://www.crescendo.ai/news/latest-vc-investment-deals-in-ai-startups
95% of Insurer AI Pilots Never Reach Production — Earnix Launches Fix
FinTech Global · Strategy
InsurTech platform Earnix has launched an AI Orchestration System (AIOS) targeting the core problem plaguing insurance AI: fragmentation, where separate tools for pricing, underwriting, claims, and customer service fail to communicate, leaving the vast majority of pilots stranded before production. AIOS combines predictive, generative, and agentic AI into a single governed decisioning layer that integrates with legacy systems via open interfaces — with every decision carrying a full audit trail of which data was used and how the conclusion was reached. For insurance executives under pressure from rising claims costs and premium stagnation, this audit-trail-first architecture is also designed to satisfy incoming EU AI Act transparency obligations.
https://fintech.global/2026/07/01/why-ai-pilots-are-failing-insurers-and-what-comes-next/