OpenAI Just Handed Its AI Agents a Credit Card

What happened: OpenAI launched ChatGPT Wallet on Friday, a payment layer that lets its “dots” agents complete purchases on your behalf — without ever seeing your actual card number. Built in partnership with American Express, Mastercard and Visa, it’s rolling out first to ChatGPT Pro subscribers in the U.S. who hold eligible U.S.-issued cards. You add a card once, review each purchase (items, total, card), then approve it with Face ID or a fingerprint. The network issues a payment token, so the merchant and the agent never touch your real credentials.

Why it matters: This is the missing half of the agentic-commerce story. For weeks we’ve watched the friction build — United banned agents from booking flights, Amazon blocked Muse from checking out. Those moves made sense because agents had no sanctioned way to pay. OpenAI is now shipping that sanctioned way, and it did the one thing that actually unlocks merchant acceptance: it made the agent’s money indistinguishable from ordinary card payments. A merchant’s existing checkout just works, rewards and purchase protections stay intact, and OpenAI published a Commerce Partner Playbook telling businesses how to stop accidentally blocking its agents.

What’s next: The telling detail is the trust ladder. Today every purchase still requires human approval — OpenAI explicitly says consumers are comfortable delegating research but not payment choice. The plan is to move toward pre-approved, parameter-based delegation over time. Watch for two things: how quickly merchants who currently block agents reverse course once “agent money” looks like normal money, and whether the big card networks’ involvement becomes the de facto standard other agent builders have to match. The wallet isn’t just a feature; it’s the infrastructure that decides whether agents browse or actually buy.

Source: PYMNTS — OpenAI Gives AI Agents a Wallet

Quick Hits

Meta’s Muse now sends 59% of all AI-agent traffic. New data from HUMAN Security shows Meta and OpenAI agents made nearly 90% of agent requests on protected sites in October, with Muse alone at 59% — 3.7x the next-largest agent, ChatGPT Agent. Agent requests are up roughly 8x since August, but 85% still hit product and search pages while checkout hovers near 2%. The takeaway: agents are learning to browse long before they learn to buy, and standalone agent browsers are losing share to agents embedded in platforms people already use. Source: Search Engine Journal

NiCE Cognigy bets the hard part of agents is running them, not building them. The contact-center platform’s next-gen release adds turn-by-turn tracing, reusable “skills,” and native voice-to-voice support — with one striking stat: AI can now run ~16 hours of autonomous work at a 50% success rate, versus minutes just two years ago. Its thesis is that “capable and dependable are not the same thing,” so the platform’s value is the feedback loop for what you build. Source: SiliconANGLE

AI News

Jensen Huang’s exit blows a hole in AI startup funding. Nvidia participated in 105 funding rounds this year — often as the largest investor — backing everyone from CoreWeave and Crusoe to Mistral AI and a $13B Hugging Face acquisition. With Huang gone and interim CEO Colette Kress expected to be far more conservative, VCs warn the early-stage infrastructure ecosystem loses one of its biggest champions, and “no one can fill” that gap. Source: TechCrunch

GrubMarket turns invoices into purchase orders with a new AI agent. The food-supply-chain SaaS provider launched its “AI Orders – Purchase Order Agent,” aiming to strip manual data entry out of fresh-food distribution. Observers note the hard part isn’t reading an invoice — it’s matching extracted data to a catalog and handling recalls, substitutions and vendor quirks, which is where the real error risk lives. Source: Digital Commerce 360

The Pentagon renames its chief AI office the “Super Intelligence Office.” The Defense Department’s CDAO — now the SIO — says the rebrand doesn’t change its mission, but brings it in line with a September directive to replace “AI” with “super intelligence” across federal agencies. It’s the second major agency to adopt the label this month, following the Justice Department’s new Super Intelligence Task Force. Source: MLQ.ai