Visa and Mastercard have completed pilot programs for specialized payment systems that enable AI agents to make autonomous purchases on behalf of consumers, with commercial launches planned for early 2026.
Both payment giants announced in recent months they are building what industry executives call “agentic commerce” infrastructure frameworks that allow AI chatbots to search products, compare prices, and complete transactions without users leaving conversation interfaces. Visa declared on December 17, 2024, that “2026 will be the final year consumers shop and checkout alone” after completing hundreds of secure agent-initiated test transactions.
What’s New in AI Agent Payment Technology
Visa unveiled its Trusted Agent Protocol in October 2024, developed with Cloudflare. The protocol uses cryptographic signatures to authenticate AI agents and distinguish them from malicious bots, enabling agents to signal purchasing intent and securely transmit payment credentials to merchants.
Mastercard launched its Agent Pay framework the same month, utilizing “agentic tokens” evolved versions of its tokenization technology to secure transactions. The company completed its U.S. rollout to all issuers in November 2025, with global expansion planned for early 2026.
T. Ramachandran, Visa’s head of products for Asia Pacific, indicated commercial deployment could begin as early as Q1 2026. Both companies are collaborating with Microsoft, Stripe, and OpenAI to develop interoperable standards.
Why This Matters for Consumers and Commerce
The infrastructure push responds directly to AI platforms integrating shopping features into chat interfaces. OpenAI launched “Instant Checkout” in September, allowing U.S. users to purchase from Etsy sellers and Shopify merchants directly within ChatGPT.
This shift fundamentally changes e-commerce by removing traditional checkout friction. Instead of navigating multiple websites and payment forms, consumers can request purchases conversationally: “Buy me running shoes under $100” or “Book a dinner reservation for Friday.”
For merchants, agentic commerce creates new discovery channels but requires integration with payment protocols to accept agent-initiated transactions. Early adoption could provide competitive advantages as AI assistants become primary shopping interfaces.
How AI Agent Payments Work
The technology enables three core capabilities:
- Agent authentication: Cryptographic signatures verify legitimate AI platforms versus fraudulent bots
- Secure credential transmission: Payment tokens transfer directly between agents and merchants without exposing card numbers
- Permission controls: Users set spending limits, merchant restrictions, and transaction approval requirements
Both Visa and Mastercard emphasize that consumers maintain complete control over agent permissions. Users can revoke access, adjust limits, or require manual approval for purchases exceeding specified amounts.
Traditional payment disputes involve four parties: consumers, issuing banks, acquiring banks, and merchants. Agentic commerce introduces AI platforms as a fifth participant, creating complex liability questions when agents make errors purchasing wrong items, booking incorrect dates, or misinterpreting user intent.
Security Challenges and Liability Questions
Payment executives acknowledge significant unresolved issues around error handling and dispute resolution. “You must anticipate errors will occur and establish safeguards around that,” Ramachandran told CNBC.
Current frameworks don’t clearly define liability when AI agents malfunction. If an agent books a wrong flight date or purchases an incorrect product variant, determining responsibility between the AI platform, payment network, issuing bank, and merchant remains legally ambiguous.
Both companies are developing error correction mechanisms and dispute protocols specifically for agent transactions. These systems will likely mirror existing chargeback processes but adapted for automated purchase scenarios.
Security protocols focus on preventing unauthorized agent access and fraudulent transaction requests. The cryptographic authentication prevents bad actors from impersonating legitimate AI platforms to initiate fraudulent purchases.
Commercial Rollout Timeline
Visa targets Q1 2026 for personalized agent transaction deployment, following successful pilot completions with multiple partners. The company has not disclosed specific merchant or issuer participants in pilot programs.
Mastercard completed its U.S. issuer rollout in November 2025 and plans European and Asia Pacific pilot programs in early 2026, with full global expansion by mid-2026.
Both companies indicate that widespread consumer availability depends on AI platform adoption and merchant integration timelines. OpenAI’s existing Instant Checkout feature provides a preview of commercial agentic commerce, though current implementations use traditional payment methods rather than specialized agent protocols.
The infrastructure race reflects broader competition in AI-powered commerce. Whoever establishes dominant protocols could control transaction flows as shopping shifts from websites to conversational interfaces, making these 2026 launches strategically critical for payment network positioning.
Featured Snippet Boxes
What are AI agent payments?
AI agent payments enable artificial intelligence assistants to autonomously make purchases on your behalf through specialized payment protocols developed by Visa and Mastercard. Users set spending limits and permissions while AI handles product search, comparison, and checkout.
When will AI shopping agents become available?
Visa and Mastercard plan commercial launches in Q1 2026, following completed pilot programs in late 2024 and 2025. Availability depends on merchant integration and AI platform adoption of the new payment protocols.
Are AI agent payments secure?
Both Visa’s Trusted Agent Protocol and Mastercard’s Agent Pay use cryptographic authentication and tokenization to verify legitimate AI platforms and protect payment credentials. Users maintain control through spending limits and merchant restrictions, though liability questions remain unresolved.
Who is liable if an AI agent makes a wrong purchase?
Liability frameworks are still being developed. Traditional payment disputes involve consumers, banks, and merchants, but agentic commerce adds AI platforms as a fifth party. Payment networks are establishing error correction mechanisms and dispute protocols for agent transactions.
