Commercial transactions typically have two human parties. A buyer initiates a purchase. A seller fulfills it. That two-party relationship is the foundation on which centuries of contract law, consumer protection legislation and payment infrastructure have been built.
Agentic commerce inserts another party into that relationship. An autonomous AI agent that researches, compares and completes purchases on behalf of the buyer, without requiring human approval at each step. The consumer sets the parameters: a budget, a preference, a deadline. The agent executes. The merchant receives an order.
But when that agent makes the wrong call, the legal and regulatory framework built for a world of human buyers and sellers is left dumbfounded regarding who is responsible. This is the chargeback liability gap.
The infrastructure driving this shift is real and already live. Google launched its Universal Commerce Protocol at the National Retail Federation conference in January 2026, backed by Walmart, Target, Shopify and more than twenty partners. Mastercard’s Agent Pay is rolling out across global markets now. AI agents are not a future concept. They are processing transactions today.
What does not yet exist is any clear framework for who is responsible when they get it wrong.
Shifting to a World of Autonomous Agents
Consumer protection law was built for a world of human and human-controlled actors. The merchant is responsible for what was sold. The consumer is responsible for the purchases they authorize. The payment system sits between both, governed by rules written for human-controlled actors on either side of a transaction.
When an autonomous machine actor enters that relationship, those rules stop working. The consumer may not have authorized the specific transaction. The merchant did not deal with a human who consented to the purchase. And when the automated AI-based agent makes a judgment call the consumer instructing the agent did not specifically authorize, the legal framework built for a human world has no clean answer for what happens next.
Failure Has No Owners
When an AI agent makes an unauthorized or harmful purchase, there are at least three parties potentially responsible with no clear liability destination among them.
The consumer delegated authority to the agent, but arguably did not authorize this specific action. The AI provider built the reasoning engine that made the call, but did not initiate the transaction directly. The merchant accepted the order, but had no way to verify the intent behind it.
Three parties potentially to blame. None of them a slam dunk case.
This is not a hypothetical edge case. It is the default state of agentic commerce as it exists today. If you want a flavor of what the failure mode can look like, a recent LinkedIn post describes a McDonald’s drive-through AI that allegedly interpreted “no” and “stop” as additions to an order rather than corrections, resulting in a spiraling bulk purchase. The story may or may not be accurate, but the underlying dynamic is real. Scale that logic to a $50,000 enterprise software contract or a bulk inventory order and the liability question becomes very serious very quickly.
The Law Is Still Catching Up to the Chargeback Liability Gap
As of 2026, no jurisdiction has enacted regulation specifically addressing autonomous AI purchasing. The frameworks that exist were not designed for this moment.
The EU AI Act, the most ambitious AI regulation to date, predates the widespread deployment of autonomous purchasing agents and contains no specific provisions for them. PSD3, the EU’s forthcoming update to its payments regulation framework, could begin to address how liability works for AI-initiated transactions, but it is still being negotiated.
In the meantime, some of the most concrete responses have come not from regulators but from the payments industry itself. Visa’s Trusted Agent Protocol and Mastercard’s Agent Pay both attempt to establish identity verification frameworks for AI agents. These are meaningful steps, but not end-state solutions.
What the Chargeback Liability Gap Means for Disputes
The mechanism for resolving liability disputes caused by agentic commerce is the chargeback; however, it was originally designed for a world in which a human initiated the transaction. When the entity that initiated the transaction is a piece of software, the standard dispute workflow has no clean path to follow.
Chargeback rules reference authorization. But what does authorization mean when the consumer authorized an agent to act on their behalf, and the agent acted outside the consumer’s intent? It is the focus on intent that will matter for resolving both legitimate third-party fraud chargebacks and first-party misuse chargebacks, AKA friendly fraud.
New data points will have to be gathered to define the scope of the cardholding consumer’s intent and whether it was exceeded or not by the agent. New regulations will have to provide clearer guidelines regarding which party is liable under what circumstances.
What comes next
Agentic commerce is projected to range from $3 trillion to $5 trillion globally by 2030, according to McKinsey research. The transactions are happening. The chargeback liability gap is real.
At that scale, the absence of a clear liability framework for agentic commerce chargebacks is not an inconvenience. It is a structural risk to the entire system.
The businesses, payment providers and dispute resolution platforms that build clear frameworks for agent-initiated transactions now, before regulation mandates them, will define the trust infrastructure that the agentic economy runs on.
