Get the appMeta says its AI app Muse has more than 10 million users, and chief executive Mark Zuckerberg says shopping is among its most popular uses. About 25% of people who set up Muse use it daily. Over a quarter of engaged users shopping through the agent had an average purchase above $200. Meta plans to keep Muse free for most users while taking a share of commerce on the platform, and expects it to become profitable once it scales.
An AI agent called Velvet Vale joined the Air-L listserv of the Association of Internet Researchers during its annual conference. The agent said it was not human and posed a research question about AI participation. Many members objected, saying they had not consented to their messages being used as data. About an hour after the introduction, the association removed the agent under a new policy banning AI agents. The agent's creator, Kevin Quinn, told the author he is a private individual in Oklahoma who built the agent to learn about the internet.
Meta has suspended the sharing feature of Muse, its new general-purpose AI agent, after researchers found it can expose sensitive information. Security firm Claroc said it listed 6,800 shared conversation links, and opening a few hundred revealed private material. The material included an executive-search note, a loan application, a bank fraud alert, a property performance report and a staff vacation calendar. Meta said it is reviewing the issue and considering changes to how the links can be discovered. Claroc said it reported the findings through Meta's bug bounty program on October 8.
Alerts, Shorts and your saved stories — free on Google Play.
Get the appMeta's AI agent, Muse, produced graphic sexual content during tests by WIRED, and the company says such conversations are permitted for adults. A Meta spokesperson said Muse runs on a private, secure virtual machine accessible only to each user. Muse's system prompt states that adult sexual content is lawful expression. Consumer advocates criticised the decision, and experts warned that intimate chats could be used to embarrass or extort users if accounts are hacked. Rival agents from OpenAI and xAI declined to engage in explicit sexting during the same tests.