{
  "id": 15721568,
  "title": "Banks want AI agents to move money, as long as someone is accountable",
  "url": "https://urgent.news/2026/10/20/banks-want-ai-agents-to-move-money-as-long-as-someone-is-accountable",
  "topic": "ai",
  "section": "AI",
  "published": "2026-10-20T07:39:30.000Z",
  "source": {
    "name": "e27",
    "slug": "e27",
    "url": "https://e27.co/banks-want-ai-agents-to-move-money-as-long-as-someone-is-accountable-20261020/"
  },
  "original_language": "en",
  "account": "The report from UOB Global Economics & Markets Research highlights that while AI agents could potentially streamline various financial tasks, the success of agentic payments largely depends on the implementation of robust governance, accountability, and security controls. This \"bounded autonomy\" approach ensures that AI agents can only execute transactions within explicit mandates, permissions, and continuous monitoring, with decisions that can be audited afterwards.\n\nHowever, the report also emphasizes that the regulatory frameworks and standards for AI agents handling financial transactions are not well-established, posing significant challenges. Interoperability across institutions, countries, and settlement networks remains limited, and liability frameworks have yet to be settled. According to the report, solving these institutional issues requires cooperation among technology providers, banks, regulators, and policymakers.\n\nSoutheast Asia is already grappling with similar challenges, particularly in the realm of agentic AI. The Monetary Authority of Singapore has set a one-year deadline for financial firms to prepare for new AI risk rules, but even in the region's most advanced market, companies adopting agentic AI still struggle with thin audit trails. Regional tech leaders are cautious and are learning to trust AI agents, but not with production-level responsibilities. They are unlikely to allow agents to move company money without human oversight.\n\nDespite these challenges, UOB identifies corporate treasury management as a potential area where agentic AI can make a significant impact. In a mature system, AI agents could assist treasurers in managing liquidity, coordinating financial workflows, and staying within institutional limits. This approach would be particularly relevant in Southeast Asia, where mid-sized companies often deal with multiple currencies, domestic payment rails, and misaligned settlement windows. AI agents could help untangle cross-border payment costs, which currently represent a significant margin loss for these firms.\n\nThe report also highlights that the infrastructure for agentic payments is improving. Singapore's PayNow Gen 2 upgrade and the gradual linking of regional real-time payment systems provide more rails for agents to use. However, this increased complexity also means that there are more places where an agent's authority must be checked and audited.\n\nWhile banks are not the only entities exploring this technology, their approach to governance and accountability is notable. Ant International has launched full-stack AI-native tools for payments, FX, and treasury, while stablecoin companies like Circle have acquired businesses like Tazapay to move into cross-border business payments. Visa and Mastercard are also developing frameworks for agent-initiated transactions, each proposing different answers to the question of who should oversee AI agents that spend money.\n\nNevertheless, scepticism is warranted. Banks' focus on governance and accountability may stem from the fact that these aspects are inherent to their existing business models. This could lead to the limitation of agents to tasks that do not threaten existing fee income. The key principles outlined in the report, including bounded mandates, portable identity frameworks, continuous authorisation models, cryptographic verification, and trusted interoperability frameworks, are essential. However, the report lacks concrete figures, such as adoption data, estimates of losses caused by agents, and target dates for the establishment of liability frameworks. It provides a set of principles without a clear delivery plan.",
  "summary": "The pitch for agentic payments usually goes like this: software that negotiates, pays, reconciles and optimises cash while the finance team sleeps. A new report from UOB Global Economics & Markets Research is a reminder that a payment involves more than moving the money. Someone also has to answer for it. The bank’s argument is […] The post Banks want AI agents to move money, as long as someone…",
  "key_points": [],
  "editors_take": null,
  "illustration": null,
  "coverage": {
    "outlets": 1,
    "also_reported_by": []
  },
  "ai_generated": true,
  "disclaimer": "Summaries, key points and the editor’s take are written by software from other outlets’ reporting and may contain errors — always check the linked original."
}