OpenAI Adds Agentic Shopping and Child Safety Tools as Altman Weighs In on Huang’s Exit

AI Insider

Source published: Site published:

Original article · فارسی

OpenAI has added a new wallet feature to ChatGPT that lets its AI agents buy things for users. It has also introduced new tools to help fight child sexual abuse material (CSAM). Meanwhile, CEO Sam Altman said the AI industry will be fine even without Jensen Huang leading Nvidia.

The new ChatGPT Wallet lets OpenAI’s AI agents, called dots, help people choose what to buy and complete purchases for them. At first, the feature is available only to a small group of US ChatGPT Pro subscribers who are at least 18 years old and have a US-issued Visa, Mastercard or American Express card.

Users add an eligible payment card to their wallet once. They can then ask their dot to help them buy something. Before the purchase is completed, users can check the items, total cost and payment card. They must approve every purchase using a security check, such as Face ID or a fingerprint. This also creates a record of the purchase. Payments use a digital token instead of the real card details, so neither the AI agent nor the seller receives the user's actual card information. The dot automatically gives sellers the user's delivery and contact details. Users also keep their card rewards and protections, including the right to dispute charges.

OpenAI said this small launch will help it test and improve the feature before making it available to more people. The company plans to add spending limits and controls for different types of purchases and time periods. These features will let users approve certain purchases in advance so their agents can handle them more independently. OpenAI also wants to work with more retailers, although some companies, including Amazon, are blocking AI agents from their websites. To help retailers prepare for shopping through AI agents, OpenAI has published a Commerce Partner Playbook with tools and advice.

This launch comes as companies that make AI products for consumers try to find business models that work. DoorDash and Uber have reported that only a small number of orders are placed by AI agents. Meta, however, has said shopping is now one of the most common activities on its Muse agent, which has more than 10 million users. Other personal AI agents, such as Instinct, can also make purchases.

Separately, OpenAI has announced new steps to fight child sexual abuse material (CSAM). It has developed a new image-checking tool with the National Center for Missing and Exploited Children (NCMEC). The tool can find abusive images, including images created by AI, and gives each image a score to help human reviewers decide what to check. OpenAI is giving selected companies free access through its API and working with Tech Coalition and ROOST to share the tool. In early tests, the tool flagged 48% fewer images while still finding the same number of abusive images. The images it flagged were also twice as likely to be abusive. This means human reviewers have fewer images to examine.

ChatGPT will also show warning screens to people who try to access material that sexualises children. It will direct them to specialist support services and suspend their accounts for 24 hours. People who break the rules more than once could have their accounts permanently removed. OpenAI developed these messages with groups including Protect Children and the Lucy Faithfull Foundation, as well as Joel Scanlan from the University of Tasmania. Scanlan independently reviewed a test of the warning messages. OpenAI has also published guidance to help developers prevent, find and report CSAM. It is also giving law enforcement agencies AI-powered investigation tools, training and a field guide.

These steps follow lawsuits claiming that OpenAI's technology has harmed users, especially children. Common Sense Media has also rated the company's technology as an unacceptable risk for teenagers. Over the past year, OpenAI has released a safety plan, made its content rules stricter, adjusted its AI models and introduced a version of ChatGPT for teenagers.

Speaking at Semafor’s Silicon Valley & The World event on Monday, Altman made his first public comments about Huang leaving Nvidia. Huang's departure followed his admission of an affair with someone alleged to be a Chinese spy. Altman said he did not know enough about the situation to say whether Huang should return. However, he said he understood how difficult the experience could be, pointing to his own removal from OpenAI in 2023. Altman said it was hard to imagine the AI revolution without Huang and the company he built. He described Huang as a hero who also has flaws.

Altman said he was not worried that Nvidia would stop providing money to support AI projects. He noted that people still trust the guarantees the company makes. Nvidia recently agreed to guarantee up to $105 billion for OpenAI's data centre project in Ohio. This is one of several deals in which Nvidia helps pay for projects that then buy its own chips. Altman said demand for AI and the money it brings in are growing so quickly that the industry will be fine. However, he also recognised the important role Huang played in moving the industry forward with great ambition.

Altman's comments come as business leaders including Elon Musk and Tim Sweeney have supported a public letter asking for Huang to be brought back. Huang is also reportedly putting together a group of Silicon Valley executives to ask for his return.