Trust, security and liability top consumers’ concerns as financial institutions look to scale agentic AI adoption with tools like Mastercard’s Agent Pay.
Mastercard is working to build consumer and merchant trust simultaneously through its Agent Pay tool, Chief Digital Officer Pablo Fourez told FinAi News. Launched in April 2025, the tool allows AI agents to make secure, tokenized payments on behalf of users, who also define the parameters for the purchases, he said.
Simplifying the process is key, Fourez said.
“Agentic payments will scale when they are both easy to build and accept,” Mastercard’s Fourez said.
“That is why we are focusing on simplifying the experience for developers and merchants alike,” he said.
Mastercard’s Agent Toolkit makes it easier for developers to build and deploy agentic payment experiences by giving AI agents structured, machine readable access to Mastercard APIs, he said.
And the FI’s Agent Pay Acceptance Framework is designed to lower the barrier for merchant participation, Fourez said.
The framework “allows merchants to recognize trusted agents and accept secure, tokenized transactions with minimal operational or technical lift,” he said. “Merchants can participate in agentic commerce without rebuilding checkout flows or adding significant new infrastructure.”
Citi and U.S. Bank are early adopters of Agent Pay in the United States, Fourez said, adding that Mastercard aims to deploy the tool to the15,000 FIs it works with around the globe in 2026.
Trust issues
But it could be a long road ahead for Mastercard. Even consumers who use AI aren’t sold on agentic AI for commerce, according to Deloitte’s “Rise of agentic commerce” report, which found:
- 58% of consumers are concerned about security, data privacy or hacking;
- 57% reported concerns about AI making poor decisions, errors or unauthorized actions; and
- 39% stated reliability and accuracy concerns.
According to the August 2025 report, to build trust in agentic experiences, institutions can:
- Allow customers to override and review agentic actions;
- Provide notifications and transparency; and
- Guarantee reimbursement for AI-related errors.
ALSO LISTEN: Podcast – Reimagining payment experiences with agentic AI
Limiting the liability
Developing trust and defining liability around the deployment of AI for making purchases presents a payment hurdle, Arjun Wadwalkar, senior product manager at Global Payments, told FinAi News.
“How do you build trust with the user that the agent will make the desired payment — and who is liable when the agent steps out of its guardrails to make a transaction?” he said.
Merchants need to feel safe to deploy agentic payments to accept transactions, and adoption will be low if they think they are on the hook for chargebacks, Wadwalkar said.
Similarly, users also must be comfortable with an agent making payments on their behalf.
The industry is considering defining liability of agentic payments very clearly in order to drive trust and, in turn, adoption, Wadwalkar said.
Security by design
Fourez agrees, emphasizing that trust starts with security by design.
“Core to Mastercard Agent Pay is agentic tokens, which are dynamic digital credentials that allow AI agents to transact securely and transparently, and guided by the permissions and intent that a consumer sets.”
Every transaction is authenticated, traceable to a specific agent and protected by the same tokenization and fraud prevention technology that secures mobile and online payments today, Fourez said.
Register here by Jan. 16 for early bird pricing for the inaugural FinAi Banking Summit, taking place March 2-3 in Denver. View the full event agenda here.
