AI & Finance News — 2026-10-23 (Fri)

Overview

US banks detected unauthorized AI agents inside customer accounts, prompting a scramble to tighten security and a warning from Jamie Dimon that "there could be thousands" of such rogue agents (Axios, Reuters, Gizmodo). AI cost worries sank Wall Street, with the Nasdaq down over 2% and sinking AI stocks undercutting markets worldwide (Reuters, Boston Herald). The bubble debate sharpened, with Ken Griffin of Citadel warning of an AI bubble while Blackstone's Jon Gray said "AI is not a bubble," and the UK's FCA set out its latest thinking on AI in financial services (Yahoo Finance, Bloomberg Television, Herbert Smith Freehills Kramer).

Banking

An Axios scoop reported that unvetted AI agents are accessing US bank accounts, and Reuters, citing Axios, reported that US banks have detected unauthorized AI agents in customer accounts (Axios, Reuters). Major banks, blindsided by AI agents, are scrambling to beef up security, and Decrypt reported that bank staff are "horrified" by rogue AI agents found inside customer accounts (Gizmodo, Decrypt). Japanese coverage examined how Wall Street's major banks are responding to growing AI costs (TBS NEWS DIG) and outlined AI-first banking spanning strategy, operations and cybersecurity (Vietnam.vn).

Insurance

Meta shares slid and default insurance hit a record as AI spending alarmed investors (Reuters). Insurers are broadening AI adoption as benefits build over the long term (Risk & Insurance). Munich Re examined the future of underwriting as the integration of digital evidence, artificial intelligence and human judgment (Munich Re). Forrester predicted that new products will fill gaps as AI reshapes insurance, and AI governance took center stage in insurance policy renewals (Forrester, Law.com).

Asset management and markets

(asset managers, hedge funds, exchanges, trading) Wall Street sank as the Nasdaq fell over 2% on AI cost worries (Reuters), and sinking AI stocks undercut markets worldwide as oil climbed to its highest price since May (Boston Herald). BlackRock maintained its AI conviction amid a US stock downgrade (Chosun Ilbo), and HarbourVest Global Private Equity noted interim headwinds alongside an AI boost (marketscreener.com). Apple's cash flow machine makes it investors' favorite AI hedge (Bloomberg), while Oracle stock was caught in the AI debt selloff, which Wall Street views as a mistake (Barron's). Modera Wealth Management launched Mira, an agentic AI offering for financial advisors (FF News). Bessent dismissed systemic AI risks at a private hedge fund investor gathering (Hedgeweek).

Fintech and payments

Visa is betting on AI agents and stablecoins for B2B payments (FinTech Magazine), and Circle Internet Group put USDC into identity-checked AI payments (Yahoo Finance). Payments executives said the AI demo era is ending (PYMNTS.com), and commentary asked whether the industry is ready for the regulatory paradigm shift of agentic payments (Winston Taylor). In Japan, an AEON subsidiary is preparing an online supermarket service in which customers place orders through dialogue with AI and complete payment in the same flow, which may start as early as the beginning of next year (Jiji Press). FINNY signed Visionary Square for a pay-as-you-go AI prospecting program (InvestmentNews). 【Financial leaders and officials】 - Jamie Dimon, JPMorgan Chase CEO, warned "there could be thousands" of rogue AI agents accessing bank accounts (AOL.com). - Larry Fink, BlackRock CEO, warned that Americans' pensions and savings will fund the trillions required for AI data centers (MoneyWise.com). - Ken Griffin, Citadel CEO, said the market is "almost certainly" in an AI bubble and recalled how past bubbles created market leaders for decades, noting he wishes he had kept his Amazon stock; he also fears a harsh awakening after the euphoria and warned of a potential AI market bubble (Yahoo Finance, Cointribune, ForkLog). - Jon Gray of Blackstone said "AI is not a bubble" (Bloomberg Television). - Adena Friedman, Nasdaq CEO and Chair, discussed the state of the AI trade (CNBC Television). 【AI-related stocks and fundraising】 (price moves, funding amounts, valuations) Chinese AI startup DeepSeek raised $15 billion ahead of a potential IPO (WSJ). AI startup Town expanded its San Francisco footprint as it hit a $1 billion valuation (The Business Journals). Venture capitalists are backing the same startup at distinct valuations amid the AI frenzy (Bloomberg), and new AI threats are fueling VC investment in cybersecurity (PEI Private Equity). A data center backlash is raising election risks for AI stocks (The Journal Record). Commentary asked what could trigger a correction in AI stocks (ConsumerAffairs), and Deutsche Bank selected three stocks as promising on the back of expanding AI use (Investing.com). 【Financial regulation and supervision】 (SEC, FCA, FSB, Japan's Financial Services Agency) The FCA published examples of good and poor practice and set out its latest thinking on AI in financial services, with its evolving approach described as running from good and poor practice to agentic finance (Herbert Smith Freehills Kramer, Freshfields, Deloitte). Compliance providers noted that regulators have made AI agent governance part of financial firms' job (Elliptic). Smarsh expanded AI and cyber risk solutions to strengthen compliance for financial firms (FF News). No SEC or Financial Services Agency (Japan) items appeared in this window. 【Integrated analysis: what is happening】 Across banks, payments and regulation, the same shift appears to be underway: AI agents are becoming actors that touch customer money directly, and the financial system is reacting on three fronts at once. Banks found unvetted AI agents already inside customer accounts (Axios, Reuters), payments firms such as Visa and Circle are building for agentic payments (FinTech Magazine, Yahoo Finance), and the FCA is extending its thinking toward agentic finance (Freshfields) — security, product development and supervision appear to be converging on the same phenomenon. The risk seems to be shifting toward identity and control — who holds the keys and who authorizes an agent — which may explain why AI governance is surfacing in insurance policy renewals, compliance tooling and certifications (Law.com, FF News, Elliptic). On the funding side, anxiety centers on how AI is paid for: swelling AI costs and an AI debt selloff weighed on equities (Reuters, Barron's), Meta's default insurance hit a record (Reuters), and Fink warned that pensions and savings will fund the trillions required for data centers (MoneyWise.com), even as private money kept flowing — DeepSeek's $15 billion raise and VCs backing one startup at distinct values suggest valuation discipline may be loosening (WSJ, Bloomberg). The split between Griffin's "almost certainly" bubble warning and the dismissals by Gray and Bessent appears to mark the point where the AI trade's financing, rather than its technology, has become the market's central worry (Yahoo Finance, Bloomberg Television, Hedgeweek).

What to watch

- Whether the discovery of unauthorized AI agents in customer accounts leads to new authentication and governance standards, and how banks and other regulators receive the FCA's agentic-finance thinking. - Whether AI cost and AI debt concerns keep weighing on equities — data center funding needs, the AI debt selloff and Meta's record default insurance may keep the bubble debate live. - DeepSeek's potential IPO and further large private-market raises, which may signal where AI funding and valuations head next.