OpenAI CEO Sam Altman on Tuesday disclosed a bug that allowed some users of its popular AI chatbot ChatGPT to view messages from others.
“We had a significant issue in ChatGPT due to a bug in an open source library, for which a fix has now been released and we have just finished validating,” Altman tweeted.
“A small percentage of users were able to see the titles of other users’ conversation history.”
Some users had reported seeing messages from others as early as Monday.
One person on Reddit reported spotting previously unseen chats in the side bar, including conversations titled “Xi Jingping’s Six Principles” and “Chinese Socialism Development.”
Another person on Twitter shared a screenshot of their ChatGPT, showing chat titles from conversations they had never had.
OpenAI temporarily disabled the chatbot on Monday after the company heard about reports of other peoples’ chat histories being visible, a spokesperson told Bloomberg.
“We feel awful about this,” Altman said in a tweet Tuesday.
The issue has since been fixed. Altman said that, as a result of the fix, users would no longer be able to see chat histories for ChatGPT conversations that took place between 1 a.m. and 10 a.m. Pacific time on Monday, March 20.
ChatGPT exploded in popularity shortly after its November launch. The tool lets users type in prompts to generate human-like responses by using what is known as a large language model, powered by vast amounts of data.
People have used it to pen school essays, song lyrics, and even generate lines of code for software.
By January, ChatGPT had amassed 100 million of monthly active users only two months into its launch, making it the fastest growing consumer application in history, according to a UBS note published last week.
It took TikTok nine months to reach 100 million users, with Instagram achieving the same popularity after two and a half years.
OpenAI will not be allowed use the word “cameo” to name any products or features in its Sora app for a month after a federal judge placed a temporary restraining order for the term on the AI startup.
U.S. District Judge Eumi K. Lee granted a temporary restraining order on Monday, blocking OpenAI from using the “cameo” mark or similar words like “Kameo” or “CameoVideo” for any function related to Sora, the company’s AI-generated video app.
“We disagree with the complaint’s assertion that anyone can claim exclusive ownership over the word ‘cameo’, and we look forward to continuing to make our case to the court,” an OpenAI spokesperson told CNBC.
Lee granted the order after OpenAI was sued in October by Cameo, a platform that allows users to purchase personalized videos from celebrities. Cameo filed a trademark lawsuit against the artificial intelligence company following the launch of Sora’s “Cameo” feature, which allowed users to generate characters of themselves or others and insert them into videos.
“We are gratified by the court’s decision, which recognizes the need to protect consumers from the confusion that OpenAI has created by using the Cameo trademark,” Cameo CEO Steven Galanis said in a statement. “While the court’s order is temporary, we hope that OpenAI will agree to stop using our mark permanently to avoid any further harm to the public or Cameo.”
The order is set to expire on Dec. 22, and a hearing for whether the halt should be made permanent is scheduled for Dec. 19.
Sam Altman, chief executive officer of OpenAI Inc., during a media tour of the Stargate AI data center in Abilene, Texas, US, on Tuesday, Sept. 23, 2025.
Kyle Grillot | Bloomberg | Getty Images
OpenAI announced a new tool called “shopping research” on Monday, right as consumers will be ramping up spending ahead of the holiday season.
The startup said the tool is designed for ChatGPT users who are looking for detailed, well-researched shopping guides. The guides include top products, key differences between the products and the latest information from retailers, according to a blog.
Users will be able to tailor their guides based on their budget, what features they care about and who they are shopping for. OpenAI said it will take a couple of minutes to generate answers with shopping research, so users who are looking for simple answers like a price check can still rely on a regular ChatGPT response.
When users submit prompts to ChatGPT that say things like, “Find the quietest cordless stick vacuum for a small apartment,” or “I need a gift for my four year old niece who loves art,” they will see the shopping research tool pop up automatically, OpenAI said. The tool can also be accessed from the menu.
OpenAI has been pushing deeper into e-commerce in recent months. The company introduced a feature called Instant Checkout in September that allows users to make purchases directly from eligible merchants through ChatGPT.
Shopping research users will be able to make purchases with Instant Checkout in the future, OpenAI said on Monday.
OpenAI said its shopping research results are organic and based on publicly available retail websites, and that it will not share users’ chats with retailers. It’s possible that shopping research will make mistakes around product availability and pricing, the company said.
Shopping research is rolling out to OpenAI’s Free, Go, Plus and Pro users who are logged in to ChatGPT.
A Tesla logo outside the company’s Tilburg Factory and Delivery Center.
Karol Serewis | Getty Images
Tesla is trying to get its “FSD Supervised” technology approved for use in the Netherlands. But Dutch regulators are telling Tesla fans to stop pressuring safety authority RDW on the matter, and that their efforts will have “no influence” on the ultimate decision.
The RDW issued a statement on Monday directed at those who have been sending messages to try and get the agency to clear Tesla’s premium partially automated driving system, marketed in the U.S. as the Full Self-Driving (Supervised) option. It’s not yet available for use in the Netherlands or Europe broadly.
“We thank everyone who has already done so and would like to ask everyone not to contact us about this,” the agency said. “It takes up unnecessary time for our customer service. Moreover, this will have no influence on whether or not the planning is met. Road safety is the RDW’s top priority: admission is only possible once the safety of the system has been convincingly demonstrated.”
The regulator said it will make a decision only after Elon Musk’s company shows that the technology meets the country’s stringent vehicle safety standards. The RDW has booked a schedule allowing Tesla to demonstrate its systems, and said it could decide on authorization as early as February.
Last week, Tesla posted on X encouraging its followers to contact RDW to express their wishes to have the systems approved.
The post claimed, “RDW has committed to granting Netherlands National approval in February 2026,” adding a message to “please contact them via link below to express your excitement & thank them for making this happen as soon as possible.” Tesla said other EU countries could then follow suit.
The RDW corrected Tesla on Monday, saying in a statement on its official website, that such approval is not guaranteed and had not been promised.
Tesla didn’t immediately respond to a request for comment.
In the U.S., the National Highway Traffic Safety Administration opened an investigation into Tesla’s FSD-equipped vehicles in October following reports of widespread traffic violations tied to use of the systems.
The cars Tesla sells today, even with FSD Supervised engaged, require a human driver ready to brake or steer at any time.
For years, Musk has promised that Tesla customers would soon be able to turn their existing electric vehicles into robotaxis, capable of generating income for owners while they sleep or go on vacation, with a simple software update.
That hasn’t happened yet, and Tesla has since informed owners that future upgrades will require new hardware as well as software releases.
Tesla is testing a Robotaxi-brand ride-hailing service in Texas and elsewhere, but it includes human safety drivers or supervisors on board who either conduct the drives or manually intervene as needed. Musk has said the company aims to remove human driers in Austin, Texas, by the end of 2025.