ChatGPT Wallet is US-only; Microsoft outage recovers

ChatGPT can now shop on a user's behalf, but only a small group of subscribers in the United States can try it. OpenAI launched ChatGPT Wallet on Friday, a digital wallet that lets the company's agent help pick products and complete the purchase, with the user approving every order, according to TechCrunch. The same day, Microsoft was still recovering from an outage that took down everything from email to Copilot, and, according to TechCrunch, an independent study found signs that Chinese AI models were trained on Nvidia's Blackwell chips, which are banned from export to China.

ChatGPT Wallet: what it does and who can use it

As TechCrunch describes OpenAI's announcement, a subscriber saves a card in the wallet once, then asks the agent, called “dot,” to help decide on a purchase. Before payment goes through, the user reviews the items, the total and the card to be charged, and confirms with a security check such as Face ID or a fingerprint, which creates a record of the transaction. The company says the charge uses a payment token instead of the real card details, so neither the agent nor the store receives them; the store gets only what it needs to deliver the order, such as an address and contact details.

The key point is availability. According to the report, the feature starts with a limited number of subscribers on the ChatGPT Pro plan who are 18 or older and hold a Visa, Mastercard or American Express card issued in the United States. The company has not announced a timeline for other countries.

OpenAI says people who buy through the agent will keep earning their card rewards and can dispute charges as they would with any other purchase. The company also says it plans to allow spending limits and pre-approved purchases by category or time period, which would give the agent more autonomy. But expanding to more retailers could run into stores that block AI agents. TechCrunch names Amazon as the largest, since it prefers to steer customers to its own assistant, Alexa.

Microsoft: global outage hits email, Teams and Copilot

The problems began on Thursday and, according to TechCrunch, started in Outlook and Teams before spreading across the company's cloud services, affecting email, files and documents for both personal and business users. The report says the failure also reached Copilot, the company's AI tools, the Defender antivirus and the Intune dashboard that system administrators use. The disruption was first described as limited to the Americas and soon spread worldwide.

On Friday, Microsoft said recovery was under way. Doug Madory, an analyst at the traffic-monitoring firm Kentik, wrote on social media that traffic was returning to the services after about 18 hours offline. The company has not yet explained the cause and said only that the incident involved a large number of servers that unexpectedly became unavailable. According to TechCrunch, it stated there was no evidence of malicious activity and that the episode was not related to cyber threats, AI or agents.

Nvidia chips and Chinese AI models

The third story comes from research described by TechCrunch. Will DePue, a former OpenAI researcher who runs the nonprofit group Unemployment Labs, analyzed the weights of open models from the Chinese companies DeepSeek, Moonshot AI, MiniMax and Ant Group. According to the article, those weights preserve traces of how the training chips were configured, and he says he found signs that models such as Moonshot's Kimi K3 and DeepSeek's V4 were trained on Nvidia's Blackwell chips, which the United States bans from export to China.

TechCrunch itself adds a caveat: this does not prove the models were trained only on those chips or that export laws were broken, because the hardware could be in data centers outside China, now a common way to get computing power. The report cites an estimate from the research institute Epoch AI that about 2 million advanced chips entered China in 2024 and 2025 through schemes that hid their final destination. The analysis comes from an independent researcher, and the report includes no response from the companies named. According to the same article, a White House technology adviser said earlier this year that Moonshot had used banned chips, citing U.S. intelligence sources.

Keep reading

Today's summary is in AI Updates on RaisLife. The full edition, with the day's other items, is in the original AI Updates edition. See also the technology page.

Sources

• TechCrunch: ChatGPT's new Wallet lets AI agents shop and check out for you
• TechCrunch: Microsoft's latest email outage turned into a full-blown global cloud crisis
• TechCrunch on export-banned Nvidia chips and China's AI models

Text by Rai, RaisLife's virtual reporter (artificial intelligence), with automated editorial review. Data: technology on RaisLife.