Monday, 28 September 2026

SEA Desk

Singapore calls for a UN framework on AI safeguards

Foreign Minister Vivian Balakrishnan told the UN General Assembly on September 26 that Singapore wants countries to consider a UN framework convention on AI safeguards, with common methods for assessing AI risks, testing models and reporting serious incidents. He said human beings must remain in control and accountable, and that Singapore will share its own AI governance and testing tools with other countries, joining Finland and Norway in pushing for tighter global standards.

NoteSingapore, home to MAS and the region's financial hub, is trying to shape global AI rules rather than wait for them. If a UN framework moves forward, expect its risk-assessment and incident-reporting language to feed back into how MAS and other SEA regulators supervise agentic AI in finance.

Source: Fintech News Singapore·Share

Rules & Regulators

FTC chair says companies can't blame AI agents for their own actions

FTC Chairman Andrew Ferguson told a Reuters event on September 25 that giving software the power to act does not make the software responsible when something goes wrong. He said regulators will look at the instructions, permissions and controls a company built into its AI agents rather than treating a bad outcome as the agent's fault, and that he prefers applying existing law before writing new AI-specific rules.

NoteThis lands the same message banks have been pre-empting on their own: whoever deploys the agent owns the outcome. It reinforces the case for the audit trails and permission logs banks and card networks are already building into agentic commerce.

Source: PYMNTS.com·Share

Banking & Lending

HSBC rolls out AI tool to speed up trade finance document checks

HSBC launched Smart Checking, an AI tool that extracts and classifies data from trade finance documents to standardize how they are processed, traced and reviewed. The bank handles more than 1 million documentary presentations a year, some running hundreds of pages, and says the tool is already live in the UK, Hong Kong and the UAE with a global rollout planned.

NoteTrade finance is paperwork-heavy, and checking a single presentation can take hours, so shaving that time off speeds up when exporters get paid. It is a narrow back-office task, the kind of AI deployment likely to spread faster than customer-facing agents.

Source: PYMNTS.com·Share

Deals

Goldman sees Big Tech AI spending topping Wall Street's own estimates

Goldman Sachs strategists led by Ryan Hammond forecast Amazon, Alphabet, Microsoft, Oracle and Meta will lift AI infrastructure spending 54% to $1.2 trillion in 2027, above the Street's $1.1 trillion consensus, after an estimated $800 billion this year. The bank says the outlay would be the largest technology investment cycle relative to GDP since 19th century railroad building, and that the five companies would need roughly $300 billion a year in AI revenue to break even on it.

$1.2TGoldman's 2027 hyperscaler AI capex forecast

NoteGoldman is both a forecaster and a lender into this buildout, so a number this far above consensus is a signal the debt and equity backing hyperscaler capex has room to grow further before anyone calls it overextended.

Source: PYMNTS.com·Share

★Top story: Banking & Lending

Threat researchers find a banking trojan whose phishing screens were written by AI

Group-IB says it uncovered RemControl, an Android trojan sold as malware-as-a-service that overlays fake screens on more than 30 banking apps across Italy, France, Spain, Poland, Portugal, Canada and Gulf states, streaming the victim's screen and logging keystrokes. Researchers found a verbatim AI assistant reply embedded in a live phishing page, evidence the operators, tracked as UNKK, used an AI coding assistant to build the panel and overlays by describing the project to it as a parental-control app.

NoteBanks' own fraud teams are racing to deploy AI defenses while the same tools are lowering the skill bar for the criminals building the attacks, so the AI arms race in fraud now runs on both sides of the ledger.

Source: Group-IB·Share

Banking & Lending

NatWest trials a voice AI that talks customers through their spending

NatWest Group is trialling a generative audio-visual tool, built on its own small language model, that lets Royal Bank of Scotland customers ask about their spending in natural voice or text and get interactive graphs back. Cora, NatWest's digital assistant, also gained a Fraud Triage Agent that has handled about 19,000 conversations routing customers who report suspicious card transactions to the right team.

NoteA homegrown model instead of a bought-in chatbot signals banks want to own the AI layer that sits on top of customer transaction data, not rent it from a vendor who also serves their rivals.

Source: NatWest Group·Share

SEA Desk

Singapore commits 80,000 finance workers to AI training by 2028

The Institute of Banking and Finance launched an AI Workforce Co-Lab with 23 banks, insurers and asset managers, including DBS, OCBC and UOB, that together employ about 40 percent of Singapore's financial sector workforce. They have committed to train more than 80,000 employees in AI skills through IBF-recognised programs by 2028, with more than half already trained. Deputy Prime Minister Gan Kim Yong, who chairs MAS, Singapore's central bank, announced the plan on September 24.

NoteSingapore is trying to retrain its banking workforce rather than let AI adoption outrun it, in roles like credit underwriting and fraud detection where agents already do part of the work. Whether training keeps pace with that adoption is the real test.

Sources: MAS, PressNewsAgency·Share

Banking & Lending

Nvidia launches agent guardrails with JPMorgan and Citi on board

Nvidia introduced the Open Agent Safety Platform, pairing runtime software that sets boundaries for AI agents with a hardware watchdog on its BlueField-4 chips that can quarantine one in milliseconds. JPMorganChase and Citi are among more than 100 organizations working with Nvidia on the open-source project. The software is available on GitHub.

NoteTwo of Wall Street's biggest banks are effectively outsourcing agent containment to their chip and cloud suppliers rather than building it in house, a sign of how fast banks are moving from experimenting with AI agents to deploying ones that can move money and need a kill switch.

Source: GlobeNewswire (Nvidia)·Share

★Top story: Banking & Lending

AI voice cloning scam drains $108 million from Italy's Fideuram

Fraudsters combined a spoofed WhatsApp message impersonating Intesa Sanpaolo's chief executive with an AI-cloned voice of a law firm partner to convince Fideuram's then chairman to authorize transfers to accounts in China and Hong Kong in February. Italian, Chinese and Portuguese authorities have recovered more than half of the roughly $108 million, about 95 million euros, taken; about 39.5 million euros remains missing, some of it converted into cryptocurrency.

NoteThis is a case of voice cloning beating a bank's internal controls rather than a customer's. Expect wealth managers and private banks, where large wires can move on a senior executive's say so, to add callback verification and delay rules for any instruction that arrives only by phone or chat.

Sources: Gulf News, Ansa.it·Share

★Top story: Banking & Lending

Apollo economist warns AI agents could drain banks of cheap deposits

Torsten Slok, Apollo's chief economist, asked "Is an agentic bank run coming?" in a note arguing that AI assistants could soon sweep household cash automatically out of checking accounts paying around 0.1 percent into accounts yielding 3.3 to 5 percent. If every household did that, he said, banks could lose much of the cheap deposit funding they lend from.

NoteDigital bank runs so far have needed people to panic together. An agent chasing yield needs only the rate gap, so a gap between checking and savings rates that regulators have long tolerated starts to look like a standing risk to deposit funding.

Source: Apollo·Share