Chargebacks911 Launches Indemnification for Merchants Against Chargebacks on AI Agent Purchases
Chargebacks911 now indemnifies merchants against qualifying chargebacks on AI agent purchases. We explain how it works, what remains undisclosed, where Visa, Mastercard and Amex stand on liability, and what e-commerce merchants should check.
Key Takeaways
- Chargebacks911 has launched a service that indemnifies merchants against qualifying chargebacks on purchases made through AI agents
- Card network frameworks focus on verifying legitimate agents and what was authorized, while liability for non-fraud disputes remains undecided
- Before relying on indemnification, merchants should be able to identify agent transactions, keep evidence trails, and confirm eligibility terms and pricing
From warning to indemnification: Chargebacks911 takes the next step
Chargebacks911 has launched indemnification for merchants against qualifying chargebacks on purchases by AI agents.
thepaypers.comWhen an AI agent authorized by a consumer buys a product and the consumer later disputes the purchase, who bears the loss? US-based chargeback management company Chargebacks911 answered that question with a product, announcing "AI agent purchase indemnification" on October 27, 2026. The service covers merchants' losses from qualifying chargebacks on agent-initiated purchases.
This is not the first time the company has raised the issue. In February 2026 it warned of a new wave of "my AI bought it, not me" disputes, and in May it pointed out again that while card networks were switching on agent payments, dispute infrastructure had not kept up. Its recommendations at the time were threefold: document authorization at the point of delegation, capture evidence across the whole transaction journey, and review fraud rules. All of these were steps merchants had to take themselves. This time the company has moved to taking on the risk itself.
How it works: identify, assess, indemnify
The service can be understood in three stages. First, it identifies which transactions involve an AI agent. Next, it assesses whether a transaction is eligible for protection. Finally, if a chargeback occurs on an eligible transaction, it covers the loss.
According to The Paypers, the assessment draws on the company's AI technology, behavioural intelligence and accumulated dispute data to read purchasing patterns, evaluate agent-involved transactions and decide which are covered. In May the company stated that it supports clients in 87 countries and processes more than 2.4 billion transactions a year through its UDMS and ResolveLab platforms. Dispute data at that scale is what underpins its coverage decisions.
For merchants, the big draw is how light the integration is. The service runs as an add-on alongside existing fraud filters and payment infrastructure, so merchants do not need to replace their fraud prevention or redesign their checkout. CTO Donald Kossmann said that in agentic commerce the transaction is only one part of the picture, and that assessing risk requires understanding behaviour, authorization and what happens after the purchase.
Founder and CEO Monica Eaton said agentic commerce should be an opportunity for merchants, not a category of transactions they must accept at their own risk. The approach is close to insurance. Instead of a binary choice between accepting or declining agent orders, it offers a third option: accept them and transfer the loss.
Many of the terms needed to evaluate it, however, have not been made public.
| Item | What has been disclosed |
|---|---|
| Coverage | Qualifying chargebacks arising from purchases involving AI agents |
| Basis for assessment | The company's AI technology, behavioural intelligence and dispute data |
| Deployment | An add-on alongside existing fraud filters and payment infrastructure; no checkout changes required |
| Pricing | Not disclosed |
| Detailed eligibility criteria | Not disclosed |
| Coverage limits and regions | Not disclosed |
The scope of "qualifying" chargebacks is the most important thing to confirm before signing. Whether coverage is limited to fraud or also includes non-fraud claims such as "this isn't what I asked for" makes a large difference to its value for merchants.
Why merchants need indemnification
The underlying reason is a gap in card network rules. Reshmi Suresh, Head of Agentic Commerce at Worldpay, notes that although Visa, Mastercard and American Express have each launched frameworks for agent transactions, liability allocation is largely unresolved once a dispute moves past straightforward fraud.
In her view, fraud cases are relatively clear. When an agent authenticates properly and a token is issued correctly, existing rules for authenticated tokenized transactions apply and the issuer carries the fraud risk. The problem is everything short of fraud. Claims that "the agent misunderstood my instruction", "it bought the wrong item" or "it went beyond what I authorized" are dispute claims rather than fraud claims, and there is no answer yet for who absorbs them.
Regulation has not caught up either. Regulation E, the US rule governing consumers' rights to dispute electronic fund transfers, still assumes a binary of authorized or unauthorized. According to Worldpay, there is no framework yet for "I asked my agent to buy something inexpensive, and it bought something expensive instead." EMVCo, which maintains the global card payment specifications, has set up a task force on agentic payments, but the article stresses that no liability shift exists yet.
Overall chargeback volume is also rising. A Datos Insights study published by Mastercard projects that global chargebacks will grow 24% from 2025 to 2028, reaching 324 million transactions a year. No one has yet shown how much agent transactions will add to that base.
Lining up the initiatives by who they protect
Sorting the initiatives of the past year or so by whom they protect makes Chargebacks911's position clear.
| Initiative | Announced | Primarily protects | Role |
|---|---|---|---|
| Visa Trusted Agent Protocol | October 2025 | Merchants | Verifies legitimate agents through cryptographic signatures |
| Mastercard Verifiable Intent | March 2026 | All parties to the transaction | Records what the consumer authorized in a tamper-proof form |
| Amex Agent Purchase Protection | April 2026 | Card Members | Protects members from charges caused by registered agent error (terms to follow) |
| Socure Agent Intelligence | October 2026 | Businesses and merchants | Detects agents and binds them to the verified person behind them |
| Chargebacks911 AI agent purchase indemnification | October 2026 | Merchants | Absorbs losses from qualifying chargebacks |
Visa's Trusted Agent Protocol has approved agents send a cryptographic signature that merchants check against Visa's registry to confirm the agent is legitimate. According to VentureBeat, the signature carries agent intent, consumer recognition data and optional payment information. Mastercard's Verifiable Intent records what the consumer authorized in a tamper-proof form. Both add evidence that can be used in a dispute, but neither decides who bears the loss.
American Express moved first on protection. Its Amex Agent Purchase Protection, announced in April 2026, commits to protecting Card Members from charges related to agent error when a registered agent the member authorized sends authenticated purchase intent (terms to follow). But it protects Card Members. Worldpay points out that while this covers members, it does not resolve what merchants absorb in disputes over agent behaviour.
Chargebacks911's indemnification targets that open merchant side. On the same day, October 27, identity company Socure announced Agent Intelligence, which detects agents and binds them to the verified person behind them. The identification layer and the indemnification layer arrived as products almost simultaneously.
Caveats
Indemnification is one company's product and does not change card network rules. Chargeback procedures, dispute ratio monitoring and network merchant monitoring all continue as before. Even if losses are reimbursed, a rising chargeback ratio still affects how a merchant is assessed.
It is also worth noting that the party deciding eligibility is the same party paying out. Because the company decides which transactions are covered, losses on transactions it deems ineligible stay with the merchant. With pricing undisclosed, merchants cannot yet compare the losses they could transfer against the fees they would pay. Worldpay argues that the liability framework itself should be rebuilt so that no single party carries undue burden. Private indemnification is best seen as a bridge until such a framework exists.
What e-commerce merchants should check now
Whether or not they consider indemnification, merchants can start on three things now.
- Make it possible to identify which orders come through agents. Without that, those orders can neither be covered nor defended in a dispute
- Keep records of what was authorized at the time of purchase, the cart contents and the confirmation screen. Amex's developer kit also shows a design that uses purchase intent and cart details in dispute investigations
- When considering indemnification, confirm in writing the definition of qualifying chargebacks, the treatment of non-fraud claims, pricing, coverage limits and regions
Conclusion
The debate over agent payments is shifting from authentication mechanisms to the allocation of liability. Until card networks decide on a liability shift, private commitments such as Amex's protection for members and Chargebacks911's indemnification for merchants will fill the gap.
The next things to watch are how much of the coverage terms and pricing are made public, and what rules EMVCo and the card networks set for non-fraud disputes. As indemnification gets priced, the industry will be able to put a number on the risk of agent transactions for the first time.
