Visa and OpenAI Launch Secure AI Agent Payments

Visa and OpenAI have announced a collaboration to introduce secure payments via AI agents. The technology will enable developers and merchants to accept automated transactions within OpenAI interfaces, such as ChatGPT Shopping and Atlas. Visa is integrating its Trusted Agent Protocol with OpenAI’s products to ensure security and control over operations.

Users will be able to delegate routine purchases to AI agents by setting predefined spending limits, merchant categories, and confirmation requirements. Visa emphasizes that all transactions undergo standard authorization and fraud monitoring. However, experts note that agentic payments introduce new security challenges, including the risk of unauthorized actions by AI.

Analysts at Forrester warn that even with traditional safeguards like tokenization, agentic payments may lead to errors and fraud that are harder to detect. Visa states that additional controls, such as tracking agent credentials, are in place for corporate clients and financial institutions to prevent unauthorized actions.

Despite the technology’s potential, trust in AI agents remains low due to incidents of erroneous purchases and ecosystem vulnerabilities. For example, AI assistants may recommend fraudulent sites as legitimate. Visa and OpenAI position their solution as a step toward widespread adoption of agentic payments but acknowledge that the technology requires further development.