When an AI agent is doing the buying, the evidence you would normally use to fight a chargeback largely disappears. There is no human click trail. No behavioral session data from the cardholder. The device fingerprint belongs to the agent’s MCP server, not the buyer’s device. Meanwhile, the authorization may be a somewhat ambiguous prompt the customer typed earlier that day, delegating open-ended purchasing power to an autonomous system.
Card networks are building new infrastructure to address this. Visa’s agent tokens and Mastercard’s Agent Pay are designed to create agent-specific authentication trails. But those tools are still maturing.
Missing evidence today includes:
- Proof that the customer delegated authority
- Timing and scope of approval
- Rules or parameters that were set
- Did the agent correctly interpret the instructions
Today, merchants operating under ACP own the liability without the evidence infrastructure needed to defend against disputes. The chargeback system was built for human-initiated transactions. When an AI agent is the shopper, that chain breaks across multiple links.
The Governance Vacuum at the Centre of Agentic Commerce
The liability problem isn’t just operational. It’s structural. New research from Darwinium, which surveyed 500 senior fraud, risk and security professionals across the US and UK, found that when an AI agent-driven transaction goes wrong, there is no consensus on who should pay.
Thirty-nine percent say the AI provider should bear liability. Twenty percent say the customer. Only 14 percent say the merchant or platform. Eleven percent say the bank or payment processor. Fifteen percent support a shared model.
That fragmentation is significant. It means the liability framework that will eventually govern agentic chargebacks is still being written. American Banker, reporting on the same research, noted that the institutions best positioned to manage this transition are those building internal expertise now rather than reacting to enforcement timelines later.
The Darwinium research also found that chargeback volume is already climbing. Datos Insights projects a 24% rise in global chargeback volume between 2025 and 2028, reaching 324 million disputes annually. Agentic commerce will initially accelerate that trend before authentication standards mature enough to stabilize it.
What Merchants Should Be Doing Now
The volume of agentic transactions is still relatively small. But precedents are being set now about what constitutes valid evidence of consumer intent, how liability flows between platforms and merchants, and what authorization means when an autonomous system is the buyer. These will define how disputes are handled at scale.
Merchant payments leads should be asking three questions today:
- Which platforms are your agentic sales coming through, and what is your position relative to merchant on record (MOR) and chargeback liability on each?
- What transaction data is actually available to you from agent-initiated purchases, and how does it differ from standard eCommerce signals?
- Lastly, if a dispute arrived tomorrow from an agentic transaction, what evidence could you produce?
The merchants who can answer those questions clearly will be in a materially stronger position than those who wait for chargeback volume to force the conversation.
Justt Was Built to Solve Agentic Commerce Chargebacks
As the nature of transactions changes, so does the evidence landscape. Justt is built to find, assemble and deploy the right evidence for each dispute, understanding the transaction context and constructing the strongest possible case from what is actually available. As agentic commerce introduces new transaction types with unfamiliar signal profiles, that capability becomes even more important.
The rules governing agentic liability are being written right now. Merchants who engage with that process, and who have the right dispute infrastructure behind them, will not be caught with their pants down when the volume arrives.
To understand your liability position in agentic commerce and how Justt can help, contact us today.
