Imagine an artificial-intelligence assistant that can order takeout for dinner or buy that new pair of pants you so desperately need. It sounds nice until you consider this: The robot needs your credit-card info.
OpenAI is trying to swerve around that roadblock, but doing so won't be easy.
On Friday, OpenAI unveiled ChatGPT Wallet, a system that enables AI agents to make transactions without accessing users' card details. The wallet performs a security check before every purchase, but it also maintains users' typical credit-card rewards and transaction protections.
OpenAI built the wallet in partnership with payment and card providers American Express, Mastercard, and Visa. The AI lab said it would work with other businesses, banks, and payment networks to expand availability over time.
The race to build personal AI agents intensified this year with the launch of Muse from Meta Platforms, but such tools ask users to give up a frightening degree of autonomy to complete sensitive tasks. While many people will be reluctant to entrust agents with their healthcare or financial information, ChatGPT Wallet is OpenAI's attempt to at least solve the payment conundrum and grow agent usage.
The eventual goal: A world where users preapprove entire sets of purchases. In theory, OpenAI could reap the rewards either through an uptick in subscriptions or by taking small referral fees when its agents makes transactions.
"Over time, we expect users to be able to delegate purchases entirely to their agent by setting clear parameters -- such as spending limits, purchase categories, or time periods," the company said in a blog post.
ChatGPT Wallet will first be accessible to U.S.-based ChatGPT Pro subscribers through Dots, OpenAI's flagship AI agent product. The company unveiled Dots in September, pitching it as an autonomous assistant for businesses and professionals.
The ChatGPT Wallet announcement indicates OpenAI also views Dots as a consumer-facing agent that can compete with Muse. The consumer focus is a core difference between OpenAI and its main rival, Anthropic, which has concentrated on attracting enterprise customers. Some 1.2 billion people use the company's ChatGPT chatbot each week, with many users relying on it to make purchase decisions, OpenAI says.
The long-term shift from in-person to online payments is evidence that so-called "agentic commerce" may eventually catch on, says Youssef Squali, an analyst at Truist Securities.
"Trust and education remain the biggest gating factors to mass adoption and to revenue growth over time," Squali wrote in a research note last month. "Trust in particular is paramount both in the model's competence and in the operator's integrity (for privacy and security)."
Retailers will have to play ball, too. Allowing an army of agent shoppers endangers retailers' direct human traffic. Within weeks of Muse's launch earlier this year, Amazon.com blocked the Meta agents from accessing its site.
Other sites, OpenAI says, may inadvertently block purchases by AI agents. The company published a "playbook" on Friday to help retailers adapt to agentic shopping.
Comments