Fintech News HK published a video that explored so called agentic commerce, highlighting that the next step in e commerce and payments was for AI systems to make transactions on users' behalf and that no stakeholder had yet solved the resulting trust problem.
The piece presented agentic commerce as an extension of digital shopping tools, where AI agents completed multi step tasks, negotiated with merchants, and initiated payments without direct human input. The video outlined three core fault lines: who retained effective control over decisions, who accepted liability for agent errors, and how financial institutions and merchants could verify the provenance and intent of automated transactions.
TRUST AND LIABILITY CHALLENGES
The video described trust as multilayered, encompassing authentication of the customer, verifiable intent, and transaction provenance. Traditional payment rails and merchant systems were built around human initiated authorisation, where customer intent could be established via interaction, physical presence, or standard digital authentication. Agentic commerce disrupted those assumptions, because a payment could originate from a decision made by software rather than a clearly attributable human action.
That raised immediate questions for banks and merchants. Authentication mechanisms could show that a device or application approved a payment, but they might not capture whether the user had actually authorised the specific action or whether the agent acted beyond its remit. Merchants could face disputes when an AI negotiated a purchase in a way the user did not intend. Banks could confront new forms of fraud or a surge in disputed payments that did not fit existing chargeback and liability rules. Regulators might view these developments as gaps in consumer protection frameworks that assume a human actor at the point of sale.
The concept of non repudiation was central to the problem. Financial institutions and payment networks rely on records that tie transactions to a payer. The video argued that agentic commerce required new forms of attestations, cryptographic proofs, or metadata standards to record the decision path of an agent, and to show whether the agent operated within limits set by the user. Without such evidence, the allocation of liability between consumer, merchant, and bank would remain contested, increasing operational costs and legal uncertainty.
IMPLICATIONS FOR BANKS, MERCHANTS AND REGULATORS
Banks, merchants and payment providers were likely to face practical changes to risk management and compliance if agentic commerce gained traction. Banks could need new monitoring rules to detect abnormal patterns that signal unintended agent behaviour. Merchants could require richer transaction metadata to assess whether a sale reflected legitimate customer intent. Payment processors and networks might need to agree on standards that attach an auditable trail to each automated decision.
For regulators, the issues cut across data protection, consumer protection, liability law and anti fraud regimes. The video framed the problem as partly technical, partly legal, and partly commercial. Without industry level agreements or regulatory guidance, the rollout of agentic payments risked fragmentation, where different providers adopted incompatible methods to record agent decisions. That fragmentation could hamper cross border commerce and create uneven protections for consumers in different jurisdictions.
Industry responses would likely require coordinated work on standards, from how to encode agent permissions to how to record and transfer decision evidence alongside payment messages. The video suggested that trust in agentic commerce would depend on interoperable mechanisms that let banks and merchants validate not only the authenticity of a transaction, but the chain of decisions that led to it, and the constraints placed on an agent by its user.
The Fintech News HK piece did not offer a definitive solution, but it underscored a practical reality for Asian financial markets and payments providers: as AI capabilities moved from recommendation to action, core payments infrastructure and legal frameworks had to adapt or risk a surge in disputes and risk exposures.
Sources: Fintech News HK