OpenAI has officially launched ChatGPT Wallet, an integrated payment tool developed in partnership with American Express, Mastercard, and Visa that allows artificial intelligence agents to complete checkouts on behalf of users. According to an announcement on OpenAIs official blog, the feature transitions AI assistants from simply comparing product listings and filling digital shopping baskets to autonomously executing external transactions. When users assign a shopping task, the agent completes the transaction through external merchant systems using integrated payment card networks. Initially, the service is accessible to ChatGPT Pro subscribers aged 18 and older in the United States, operating within the dots environment rather than regular chat windows, Work, or standalone Codex threads.

Unlike traditional e-commerce flows that require buyers to manually type a 16-digit card number, expiration date, and security code or follow separate payment links, ChatGPT Wallet requires only a single card registration. When executing a purchase, the agent generates virtual credentials at the checkout window, ensuring actual card numbers remain hidden from both the AI and the seller. Transactions rely on single-use network tokens, dynamic expiration dates, and temporary security codes, while preserving cardholder rewards points and existing dispute protections. To prevent unauthorized or unintended purchases, the checkout requires human verification: users must review the itemized list, total charge, and designated card before finalizing the transaction through device-based biometric authentication such as Face ID or fingerprint scanning.

Despite these safeguards, agent-driven commerce faces resistance from major retail platforms and infrastructure bottlenecks. Global brands including Amazon and United Airlines have deployed bot countermeasures that restrict automated AI agents from entering systems or booking flights, citing internal security policies and transaction integrity. Furthermore, standard anti-fraud detection systems frequently misidentify proxy agent checkouts as anomalous traffic, blocking access to checkout screens. Security analysts note that even with tokenized card data, risks such as prompt injection attacks or redirection to counterfeit storefronts remain unresolved. Establishing broader merchant compatibility and refining anti-fraud detection will serve as key benchmarks for whether autonomous agentic commerce can expand globally.