Welcome to the November 8, 2026 edition of Money Explored—the essential Sunday briefing to stay ahead in fintech.
AI agents and fast-growing fintechs have been moving money faster than anyone could check it. This week a regulator, a bank and federal prosecutors each said who answers when it goes wrong.
Three signals. One direction: whoever lets money move fast now has to prove someone stayed in control.
THIS WEEK:
OCC's AI playbook: Examiners will judge an AI agent by the damage it could do, not how clever it is.
Wells Fargo's agent terms: Customers must prove they switched their AI agents off.
Block's $764m DOJ deal: Cash App's weak money-laundering checks get a price.
Plus: Hong Kong starts writing rules for AI agents, Thailand moves to block Bitget, Revolut hits $115bn, and a Japanese giant backs a US bank built for stablecoins.
This edition is published in partnership with Attio. See how revenue teams are getting AI agents to review sales calls and draft follow-ups in minutes, not half an hour, below.
🌎 Major Stories
The Office of the Comptroller of the Currency, the regulator that charters national banks in the US, published its first framework for examining how banks use AI on 4 November. It calls the technology "super intelligence" and covers agents that act on their own. Examiners are told to judge each use by what happens if it fails, how much power it holds, and how closely a human watches it. An agent that can move money draws more scrutiny than a tool that only drafts text. The OCC says the framework guides its examiners and sets no new rules for banks.
Strategic Takeaway: This fills the gap left in April, when the bank regulators' new model-risk guidance put generative and agentic AI out of scope. The OCC is not writing a separate AI rulebook. It folds agents into the checks banks already face, scaled to the harm an agent could cause. It also says banks can limit what a vendor's agent does inside their systems, and use contracts to decide who pays when it gets things wrong. For builders selling AI to banks: expect your bank client to ask what your agent can do without a human, and who covers the loss when it errs.
Wells Fargo has updated its online banking terms, effective 24 November. A new section covers AI agents that customers let into their accounts. Customers stay responsible for what an agent does until they tell the bank it has been switched off, where the law allows. They must also keep proof that they revoked the agent and hand it over if the bank asks. American Banker reports that Wells Fargo appears to be the first bank to go this far.
Strategic Takeaway: Big banks now have two answers to AI agents. JPMorgan, Bank of America and Citi have blocked them from customer accounts, Banking Dive reports. Wells Fargo lets them in but puts the risk on the customer. The terms take effect three days before Black Friday, just as shoppers start handing agents their logins. For operators: if your product acts inside a customer's bank account, check whether the bank's terms leave your users carrying losses your agent causes.
Block, the owner of Cash App and Square, has reached a deal in principle with the US Department of Justice. The DOJ says Block failed to keep an effective anti-money-laundering program, the checks that stop criminals moving money through an app, as the Bank Secrecy Act requires. Block disclosed the deal in its quarterly filing on 3 November and has set aside $764m. The DOJ would hold the charge for three years, then drop it if Block meets the deal's terms. The deal is not final, and Block warns the cost could rise.
Strategic Takeaway: The probe began after a 2023 short-seller report on Cash App's checks. In July, Block paid $45m to settle state claims that Cash App failed to protect users from fraud. The DOJ deal landed the same day Block said Cash App's gross profit grew 27%. Fast, easy sign-up built the business, and the bill for its weak checks is now in the hundreds of millions. For funders: when you value a consumer fintech, ask how much of its growth came from light identity checks, and what fixing them would cost.
SPONSORED CONTENT: ATTIO
Frontier models perform better in Attio’s agent harness.
We gave GPT-6 Astra the same job twice: analyze 100 sales calls and draft the follow-ups, once in Ask Attio and once in ChatGPT.
Ask Attio finished in 4 minutes 49 seconds. ChatGPT took 26 minutes 11 seconds, about 5.4 times as long.
The most ambitious GTM work requires more than frontier intelligence. It requires architecture that can put that intelligence to work at scale.
