{
  "id": 16880910,
  "title": "Who is Responsible When an AI Agent Makes a Payment? LaFinteca CEO Dmytro Rukin eEplains",
  "url": "https://urgent.news/2026/10/28/who-is-responsible-when-an-ai-agent-makes-a-payment-lafinteca-ceo",
  "topic": "ai",
  "section": "AI",
  "published": "2026-10-28T09:54:53.000Z",
  "source": {
    "name": "HackerNoon",
    "slug": "hackernoon",
    "url": "https://hackernoon.com/who-is-responsible-when-an-ai-agent-makes-a-payment-lafinteca-ceo-dmytro-rukin-eeplains?source=rss"
  },
  "original_language": "en",
  "account": "When an AI agent makes a payment, responsibility lies in following the mandate that the user agreed to beforehand, according to Dmytro Rukin, CEO of LaFinteca. The user sets the limits, the agent provider ensures the agent stays within those limits, and the merchant and payment provider conduct only the checks that they are capable of.\n\nOn March 25, 2026, a global card network introduced live agentic payments across Latin America and the Caribbean, involving 17 financial institutions that had all obtained cardholder consent, as per its announcement. The question of who bears responsibility for a payment made without human intervention remains open. The FCA highlighted this issue in its Payments Regulatory Priorities report of the same date, while HM Treasury's consultation on July 14, 2026, asked whether consent and liability rules should be adapted.\n\nThe core question is what constitutes consent when a machine is making a payment. Rukin proposes four limits as the answer: an amount cap, a specific set of merchants, a time frame after which the permission expires, and a quick cancellation option. To ensure clarity, a stranger should be able to determine without guessing whether a given payment falls within the consented boundaries.\n\nPix Automático, a service launched by Banco Central do Brasil on June 16, 2025, exemplifies how Brazil handles agentic payments. After the payer authorizes a company once in their banking app, they can set a maximum amount and cancel at any time. Subsequent payments occur without further authentication, as stipulated by the central bank. This system was designed for monthly bills and demonstrates a potential structure for AI agent mandates.\n\nIn the UK, variable recurring payments follow a similar logic, allowing customers to provide a mandate with pre-agreed limits, review, modify, or cancel it at any time. The commercial scheme became operational on June 2, 2026, and according to the FCA, variable recurring payments accounted for roughly 16% of open banking payments as of June 2026. Rukin argues that people in London seem to overlook the fact that such consent-based systems already exist.\n\nRukin, representing LaFinteca, which facilitates business payments between Latin America and Europe, proposes that the entity best positioned to prevent a mistake should bear the loss. If the agent remained within the established limits, the user is responsible. If the agent provider overstepped its bounds, the provider bears the responsibility. The merchant assumes liability if it overlooked a visible limit on a payment flagged as agent-initiated, and the payment provider is liable when a payment exceeding the cap still goes through. He emphasizes that card disputes have been handled in a similar fashion for decades, with the cost falling on the person who could have prevented the error most economically. The additional layer introduced by the agent merely adds another participant to this process.\n\nTo prepare for this emerging trend, Rukin suggests four steps for merchants and payment teams in the current quarter. First, record each mandate as easily readable fields that the customer can verify. Second, maintain a timestamped log detailing what the agent was authorized to do and the reason for each transaction. Third, streamline refunds into a two-step process. Finally, ensure that the agent-initiated flag, which card networks already transmit in authorization data, is incorporated into order and refund records. Rukin firmly believes that trust in payments hinges on consent and pre-defined limits, and these principles must be applied to AI agent payments as well.",
  "summary": "Who is liable when an AI agent pays? LaFinteca CEO Dmytro Rukin on consent, VRP, Pix Automático and who should carry the loss in agentic payments.",
  "key_points": [
    "Responsibility lies in user's pre-set limits, not AI agent's actions",
    "Global card network introduced live agentic payments March 25, 2026",
    "Rukin proposes four limits for consent: amount cap, merchant set, time frame, cancellation option"
  ],
  "editors_take": "Shifting liability to the party best positioned to prevent errors transforms AI agent payments into a structured extension of existing consent-based frameworks used in traditional banking systems.",
  "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."
}