Visa Embeds Payment Network into ChatGPT — Agentic Commerce Goes Live
View original source →OpenAI and Visa announced on June 11 an integration enabling ChatGPT AI agents to execute purchases on a user's behalf through Visa's tokenized payment network, with anti-fraud protection running through Visa's existing transaction rails.
Key features:
• Unique payment tokens are created for each AI agent session — the agent never handles actual card numbers • Users define spending parameters: maximum amounts, approved merchant categories, blacklisted vendors • Transactions outside parameters require explicit human approval • Initial coverage includes physical goods, digital goods, hotels, flights, and software subscriptions • Visa's partnership is non-exclusive — discussions are active with Anthropic, Google, and Microsoft
The Consumer Financial Protection Bureau and the UK's FCA both issued statements noting they are monitoring agentic commerce developments and will evaluate whether existing consumer protection regulations adequately cover AI-authorized transactions.
No major jurisdiction currently has regulations specifically addressing AI agent autonomous purchasing authority, consumer recourse for AI-authorized transactions, or liability allocation when an AI agent makes an unintended purchase. Organizations deploying AI agents with payment capabilities are operating in an unregulated space.
Why It Matters: The infrastructure for AI agents that can reason, authenticate, execute transactions, and complete purchases without human approval is now operational. Governance frameworks have not caught up.