Tuesday, 29 September 2026

★Top story: Deals

Anthropic's IPO prospectus discloses billions in losses and an AI risk warning

Anthropic's prospectus for its planned US listing shows a 2025 operating loss of more than $8 billion against revenue that grew twelvefold to nearly $4.6 billion, with the company targeting a valuation above $2 trillion. Reuters and the Financial Times reported that the filing is the first on the SEC's database to list existential risk to humanity as a disclosed risk factor.

$8B+Anthropic's 2025 operating loss

NoteA company does not usually put existential risk in the same document as its revenue growth, so seeing Anthropic name it as a risk factor is as much a signal to the banks building on its models as it is to the investors pricing a valuation above $2 trillion.

Source: TechCrunch·Share

Banking & Lending

Nasdaq opens a sandboxed AI agent environment inside its Calypso platform

Nasdaq built a governed operating environment inside its Calypso trading platform that lets banks run AI agents against trading, risk and collateral data with sandboxing and no external data retention. The first capability is a natural-language assistant for querying platform data and documents, and Nasdaq said nearly half of firms it surveyed pointed to manual work as a bottleneck in post-trade operations.

NoteBanks want agents working their trading desks but not with free rein over the system of record, so wrapping them in the exchange's own sandbox is a template other market infrastructure providers are likely to copy.

Source: Nasdaq·Share

Payments & Fintech

Mastercard launches AI tool to flag which suppliers will take a card

Mastercard rolled out Advanced B2B Analytics, which scores corporate suppliers on how likely they are to accept card payments by analyzing accounts payable data and shows the results in dashboards for banks and their clients. Absa Group and Emirates NBD are among the first banks offering the tool to corporate customers.

NoteB2B card payments lag consumer ones because banks rarely know which suppliers will actually say yes to a card, so a scoring tool turns that guesswork into a sales lead list.

Source: Fintech News Singapore·Share

SEA Desk

Philippine anti-money laundering body tells banks to prove AI actually works

The Anti-Money Laundering Council's executive director, Ronel Buenaventura, said banks using AI to fight financial crime must be able to explain and audit how the systems work, from the data used to how decisions are checked, and show the tools actually help detect or prevent crime rather than just speed up compliance work. He said boards must receive and act on risks the technology flags, and human review stays mandatory.

NoteThe AMLC is doing what US federal banking regulators have so far avoided: setting explicit explainability and audit requirements for AI-driven compliance tools, ahead of the broader agentic AI rollout Philippine banks are only starting.

Source: Philstar.com·Share

Banking & Lending

Capital One says agentic AI lives or dies on data plumbing

Capital One vice president of enterprise AI Rashmi Shetty said the bank treats agentic AI as an end-to-end system, not a model choice, leaning on governed data pipelines, sandbox testing, observability of agent actions and human review for high-risk steps. The bank's Chat Concierge multi-agent tool has been in production for more than two years helping customers shop for cars.

NoteThe lesson from a bank that has run agents in production the longest: the hard part is not picking a model but building the data lineage and guardrails around it, which is the work still missing at banks racing to bolt agents onto legacy pipelines.

Source: Banking Dive·Share

Models & Labs

Anthropic ships a cheaper Sonnet that nearly matches Opus

Anthropic released Claude Sonnet 5.5, which it says runs more than 30% faster and costs up to 30% less per task than Sonnet 5, at the same $2-per-million-input, $10-per-million-output pricing. On benchmarks such as GDPval-AA and OSWorld 2.1, Sonnet 5.5 scores within a few points of the larger Opus 5.5 model.

NoteBanks and fintechs running Claude at scale for coding, document work and agents get most of a flagship model's capability at mid-tier pricing, which lowers the cost floor for the agentic pilots already spreading across back-office and customer-service work.

Sources: TechCrunch, The Decoder·Share

Rules & Regulators

State bank regulators publish their own AI examination playbook

The Conference of State Bank Supervisors released an AI Supervisory Framework on September 16 to help state examiners review AI use at state-chartered banks and state-licensed nonbank financial institutions, covering governance, AI inventories, generative AI, third-party risk and consumer protection. CSBS president Brandon Milhorn said AI "provides a powerful new tool for financial institutions to improve services, protect consumers, and increase operating efficiency."

NoteThe OCC, Federal Reserve and FDIC updated joint model-risk guidance in April but left generative and agentic AI out of scope, calling it too new to pin down. States just filled that gap themselves, so a state-chartered bank's AI exam can now be more specific than a federally chartered peer's.

Sources: CSBS, PYMNTS.com·Share

Banking & Lending

Bank of America adds AI benchmarking to its CashPro treasury platform

Bank of America launched Payments Insights, an AI-powered addition to CashPro Data Intelligence that gives corporate and commercial clients peer benchmarking and visualizations on payment efficiency, cross-border flows and working-capital performance. The bank says CashPro handled 213 million payments in the first half of 2026, up 10% from a year earlier.

213Mpayments CashPro handled in H1 2026, up 10% year on year

NoteTreasury teams have had their own payment data for years; what they lacked was a peer benchmark to know if their process is actually slow. Folding that into an existing platform, rather than selling separate analytics, is how a large bank makes an AI feature feel like table stakes.

Source: Bank of America Newsroom·Share

Banking & Lending

Milan prosecutors find a second AI voice-cloning fraud, this time at Banca Ifis

A Banca Ifis manager authorized about 24 million euros in transfers in May after a series of fraudulent emails and a cloned voice, the same method used against Fideuram's then chairman in February, according to Ansa. About 20 million euros were recovered after the money was traced through Spain, Singapore, Hong Kong and Bahrain and seized in Barcelona; prosecutors have opened a third, smaller case worth about 2 million euros.

€24Mtransferred after the AI-assisted fraud

NoteMilan's prosecutors are now treating AI-voice fraud against bank staff as a caseload, not an isolated incident. That argues for callback verification and a cooling-off period on any large wire authorized only by phone, message or email, at any bank, not just the ones already hit.

Sources: Ansa.it, beBankers·Share

Payments & Fintech

Shopify opens checkout to browser-based AI shopping agents

Shopify extended its WebMCP standard with three new tools, get_checkout, update_checkout and complete_checkout, letting an AI agent read a cart, change details like address or delivery, and submit a purchase with the buyer's authorization instead of scraping the page. The rollout covers all eligible Shopify merchants, with Meta's Muse and Instinct AI named as early agent partners.

NoteThis gives agents a structured door into checkout instead of a screen to fake being a shopper on. It moves the agentic commerce question from whether agents can buy to who is liable when one gets the order wrong, the gap last week's bank consortium principles were trying to close.

Source: TechCrunch·Share