AI and Financial ServicesWeekly briefing · Monday, October 19, 2026Covers Oct 12–18 · 30 stories
Big banks block outside AI agents from customer accounts. Payments firms are defining who authorizes agent-led purchases.
Financial firms are moving from AI experiments to access control. The near-term winners will be banks, card networks, and insurers that can prove consent, limit automated actions, and keep people accountable for consequential decisions.
Written with AI from the 30 stories we linked this week. The small numbers point to the story list at the end.
The week in brief
- Banks restrict outside agents from direct account access.
- Payments firms build rules for agent-authorized purchases.
- Regulators focus on AI control, recovery, and consumer harm.
- Insurance keeps human judgment around pricing and claims.
- AI shows narrow workflow gains, not yet clear bankwide returns.
Banks are closing the door on outside agents
JPMorgan Chase said it will not let outside AI agents directly access retail customer accounts. Citi has also restricted agent access, while Charles Schwab is blocking agents that use client credentials or browser automation.123
The concern is not only fraud or mistaken trades. Banks and brokers also fear that an agent acting for a customer could move idle cash into higher-yielding accounts, weakening a major source of interest income.23
Several rivals are testing controlled ways to connect customers' preferred agents. Bank of America, BBVA, Synchrony, and Wells Fargo joined the Personal Agent Protocol, while Robinhood and Coinbase already permit certain links between customer accounts and AI assistants.4
So whatAccount providers are trying to keep control of the route into customer money. That favors firms with their own assistants and approved connections, while making independent consumer agents less useful in financial services.
Agent payments need proof of consent and someone to pay for errors
American Express introduced an Intent Intelligence Framework that it says carries a customer's purchase instructions through an agent-led transaction. The company is seeking to make it an industry standard.5
The unsettled issue is liability when an agent makes the wrong purchase, is manipulated, or follows a bad instruction. Federal Reserve Governor Christopher Waller said existing card-network and consumer-protection rules may be adaptable, but could require testing and refinement.6
Visa argues that standards developed for online and mobile payments can help protect agent-led payments, and Stripe President John Collison said the processor's agentic-commerce activity rose sharply in September without publishing figures.78
So whatThe first commercial fight in agent-led payments will be over authentication, chargebacks, and liability, not shopping features. Card networks and banks that set those rules can determine which agents merchants and customers are allowed to use.
Supervisors are treating AI control as a financial-stability issue
U.S. banking regulators are asking banks about advanced models, including their use of Chinese technology, AI in cyber defense, and their ability to control autonomous systems. European supervisors are pressing banks to strengthen recovery plans for AI-related disruption.9
The UK Financial Conduct Authority plans to use AI to automate parts of financial regulation.10
In consumer banking, a report by former regulators argues that banks remain liable when customer-service chatbots give wrong answers, make decisions affecting finances, or prevent customers from reaching help.11
So whatAI governance is becoming an operating cost rather than a policy exercise. Banks that cannot show how an automated action can be stopped, recovered, and explained will face slower deployment and more scrutiny.
Insurance keeps AI on a short leash while banks find narrow gains
Insurers remain reluctant to let generative AI determine rebuilding costs. Openly's technology chief said cost models need to be transparent and deterministic, while industry participants emphasized that experienced professionals should validate estimates.12
Property and casualty insurers asked the National Association of Insurance Commissioners to exclude generalized linear models, widely used in insurance pricing, from its AI evaluation tool. The NAIC plans to finalize that tool at its November 14 to 17 fall meeting.13
Colony Bank says generative-AI loan software cut its SBA lending workflow from 92 steps to 32 and reduced approvals to about seven business days for responsive customers. At the other end of the scale, BNY says it is putting AI across its business but declined to disclose its spending or quantify the returns.1415
So whatInsurance is likely to automate measurement and paperwork before judgment on prices and claims. Banks can show gains in tightly defined workflows, but broad claims of AI payback still lack the cost and return data investors want.
Caution: Company-only claims cited this week include American Express's new intent framework and its standard-setting aim, Stripe's reported rise in agentic-commerce activity, Colony Bank's reported workflow and approval-time gains, and BNY's claimed AI returns.581415
This week’s stories
All 30 stories we linked in AI and Financial Services, Oct 12–18. Headlines open the original.
Discussed above
- 1JPMorgan Chase blocks outside AI agents from directly accessing customer accounts
- 2Schwab blocks unauthorized AI agents, points clients to 'Charley'
- 3Citi becomes latest US bank to restrict AI agents from accessing accounts
- 4Why some banks are exploring 'bring your own AI agent'
- 5Amex introduces a framework to prevent AI agents from making unauthorized purchases | TechCrunch
- 6Bankers grapple with liability when AI agents mess up
- 7Should banks prepare for an 'agentic buy button''?
- 8Stripe agentic commerce revs up
- 9Banking Watchdogs Ramp Up AI Scrutiny as Stability Fears Mount, Sources Say
- 10FCA plans to embrace AI to automate financial regulation
- 11Banks face liability over use of AI chatbots, report finds
- 12AI falling short on rebuilding cost estimates for insurers
- 13Insurers: Regulators should drop pricing models from AI rules
- 14Colony Bank uses gen AI to shave days off SBA loan approvals
- 15While embedding AI 'everywhere,' BNY won't disclose costs
Also this week
- 16FIS launches agentic AI suite for the office of the CFO
- 17Keynova: Top insurer apps offer savings, policy updates—not gen AI
- 18Market data costs to rise by up to 50% thanks to AI
- 19Harbor Emerges From Stealth With $10M Funding to Handle AI-Agent Risk
- 20SBS unveils sovereign AI platform for core banking
- 21A quarter of financial firms tie pay and promotions to AI proficiency: survey
- 22NatWest enlists Nvidia to help its AI spending tool understand regional accents
- 23Hippo's helpers: How insurer holds AI agents to human standards
- 24Lack of financial confidence driving consumers to AI assistants
- 25AI-native payments platform Movere launches
- 26AI-native personal finance platform Matters emerges from stealth with $5m pre-seed
- 27Broad AI Exclusions in Insurance Policies Could Create Uncertainty, Marsh CEO Says
- 28People Moves: OneShield Appoints Brown as Chief Insurance Officer
- 29Deutsche Bank rolls out on-premise AI system
- 30M&T says it's monitoring impact of AI agents on deposits
Earlier AI and Financial Services briefings · Today’s Financial Services news