Wednesday, 23 September 2026

★Top story: Banking & Lending

Meta's Muse agent knocks bank, insurer and broker stocks

Investors sold companies that profit when customers never switch after Meta's Muse, an agent that can compare services and buy on a user's behalf, hit the top of Apple's US App Store. The S&P 500 Financials Index fell as much as 2.4% on Tuesday to its lowest since July, JPMorgan, Wells Fargo and Morgan Stanley each dropped more than 2.5%, and Allstate and Charles Schwab slid more than 5%.

−2.4%S&P 500 Financials at Tuesday's low

NoteGoldman frames this as a bet against consumer inertia: fees and add-ons that survive because switching is a hassle. An agent that compares and switches for you turns that inertia into churn, and a lot of retail banking and insurance margin sits right there.

Sources: Bloomberg, Investing.com·Share

★Top story: Banking & Lending

Six global banks set rules for AI agents that shop and pay

ASB Bank, Bank of America, Capital One, Commonwealth Bank of Australia, ING and NatWest published Building Trust in Agentic Commerce, a set of voluntary principles for AI agents that buy on a customer's behalf. They want an audit trail of what the customer asked for, what authority they granted and what happened at payment, and they flag agents that ask for card details and key them into websites.

NoteThe banks are writing the rulebook before a regulator does. If agents must identify themselves and leave a trail, banks keep their seat at the checkout, and the question of who pays for an agent's mistake gets answered before the first big loss.

Sources: PYMNTS, Finextra·Share

Payments & Fintech

Amazon blocks Meta's Muse from shopping on its store

Amazon asked to be removed from Muse's shopping experience and blocked the agent from its site, saying it was never told Muse would access the store and did not authorize it. Muse pays for purchases through Link by Stripe.

NoteThe first real fight in agentic commerce is over who stands between the customer and the checkout. Merchants now want what six banks asked for this week: agents that say who they are.

Source: Payments Dive·Share

SEA Desk

Indonesian lawmakers warn banks not to use digitalization and AI as cover for layoffs

At a hearing on Tuesday, parliament's finance commission, Komisi XI, asked OJK, Indonesia's financial regulator, to review bank governance so digital transformation isn't used to justify one-sided dismissals. The complaint came from the Maybank Indonesia workers' union, and member Amin AK asked OJK to evaluate the bank's performance improvement plan scheme, which he said had become a one-sided tool to pressure staff.

NoteThe politics of AI job cuts have reached Indonesian banking before most of the savings have. Banks planning automation now need a redeployment story ready, because that is the question legislators are asking.

Sources: DPR RI, DPR RI (Amin AK)·Share

SEA Desk

Deutsche Bank puts AI agents on wealth KYC in Singapore and Hong Kong

Deutsche Bank's private bank rolled out an agentic AI tool for source-of-wealth checks across its Singapore and Hong Kong booking centres, automating the research and paperwork behind proving where a client's money came from before staff review it. The bank estimates a 30% uplift in client onboarding.

30%estimated uplift in client onboarding

NoteSource-of-wealth checks are one of the slowest steps in Asian private-bank onboarding. Bank of Singapore says its own agent cut onboarding from more than 30 days to about 15, so this is becoming table stakes on Singapore's wealth desks.

Sources: Finextra, Private Banker International·Share

SEA Desk

Nearly half of Singaporeans would let an AI agent apply for credit, Experian finds

In an Experian study run by Forrester Consulting, 49% of 482 Singapore respondents were comfortable with an AI agent applying for a loan or credit card, and 83% trust AI to compare loans. The top worry, cited by 82%, is AI being manipulated by fake offers, the highest share across the 13 markets surveyed.

49%comfortable letting an AI agent apply for credit

Note77% would be more comfortable if the agent were linked to a trusted financial institution. That is the opening for banks: become the agent customers trust before a big tech assistant becomes the front door to credit.

Source: Fintech News Singapore·Share

Models & Labs

Anthropic and OpenAI cut frontier model prices on the same day

Anthropic released Claude Opus 5.5 at $4 per million input tokens and $20 per million output, down from $25 output on the previous Opus, and called it the strongest model it has tested. OpenAI answered with GPT-6 Sol and Luna at half the price of the models they replace, and says Sol makes about half as many mistakes as its predecessor.

50%OpenAI's price cut on GPT-6 Sol and Luna

NoteFor a bank's AI budget, price matters more than any benchmark. Statement analysis, KYC checks and document reading are high-volume jobs, and every cut like this moves more of them past the ROI line.

Sources: TechCrunch (Opus 5.5), TechCrunch (GPT-6 Sol, Luna), Tech in Asia·Share

Payments & Fintech

Mastercard and Rogers Bank complete what they call Canada's first agentic purchase

A Flybits-built AI assistant followed instructions set by a Rogers Red credit card holder, found an eligible product within a set budget and paid for it using Mastercard Agent Pay. The partners say 41% of Canadians they surveyed are open to letting AI assistants help them shop online.

NoteThis is the other half of the week's agent story. While banks write principles and Amazon blocks Muse, the card networks are building rails that let an agent pay inside limits the customer set.

Sources: Rogers, Finextra·Share

Deals

Snorkel AI raises $350 million at a $3.5 billion valuation

Snorkel AI raised a $350 million Series E led by Insight Partners and S32, lifting its valuation to $3.5 billion from $1.3 billion 17 months ago. The Stanford spinout now sells finished training datasets and simulated environments to AI labs and companies, and says its annualized revenue run rate is $375 million.

$3.5Bvaluation, up from $1.3B in 17 months

NoteThe money is following the scarce input. Labs have the compute, and what they can't scrape is expert-labeled data, which in finance means credit, fraud and compliance cases.

Source: TechCrunch·Share