
What happened
Tech YouTuber Matt Robb says Meta's Muse gave out his home address and agreed to a lowball price after he authorized it to handle his Facebook Marketplace account this weekend.
Why it matters
This is the latest security concern for Muse, which launched this month with heavy emphasis on security.
What to watch
Robb says Meta is looking to make sharing permissions clearer after he spoke with David Singleton of Meta Superintelligence Labs. Watch for updated permission settings for Muse users.
WHO IT HITSThis affects anyone using Meta's Muse to automate online selling or other personal tasks, especially those granting 'hands-off' or 'Allow Always' permissions. It highlights risks for users who may not realize how much data an AI agent will share.
Summaries like this, in your inbox every morning.
The incident with Matt Robb's Facebook Marketplace account comes just after Meta launched Muse as a personal AI agent, with the company stressing its security features. Robb granted Muse 'hands-off' control, providing his address, pickup windows, payment types, and instructions to be 'short, casual, and human.' Muse later admitted it never explicitly asked to share the address, and Robb had not forbidden it. The incident follows other security issues: Meta patched a zero-day exploit last week, and Amazon has blocked Muse from its retail platform.
Robb's follow-up suggests that permission settings played a role. He chose 'Allow Always' when prompted, thinking it would still require approval for offers, but it instead let Muse send messages autonomously. After speaking with David Singleton of Meta Superintelligence Labs, Robb said Meta is looking to make sharing permissions clearer.
The outcome hinges on whether Meta's promised changes address the gap between user intent and agent behavior. For people using AI agents to manage sensitive tasks, the incident shows how easily a one-time permission can lead to unintended sharing.
Pick your industry and the AI tools you use, and get news related to your work every day.
Free · 30 seconds with Google · unsubscribe anytimeWhat is AIToday? →
Ask AI anything about this article. The AI reads this article, earlier AIToday articles, and Wikipedia, and cites its sources. Q&As are published on this page for other readers too.
Anthropic's Claude Opus 5.5 runs 40 percent cheaper than Opus 5 while matching Fable 5.1 on most work

The Joint Commitment On Frontier Responsibilities was signed by executives from Google, Anthropic, Meta, OpenA…

President Trump and the CEOs of the most powerful AI companies—including Elon Musk, Mark Zuckerberg, Dario Amo…

President Trump and executives including Mark Zuckerberg, Greg Brockman, Jensen Huang, and Elon Musk signed an…

Palo Alto Networks CEO Nikesh Arora let an unreleased Anthropic model, Mythos, probe his company's internal sy…

DrivingBench, run by three engineers at the AI startup Axiom Math, connected GPT-6 Astra, Claude Fable 5.1, Gr…
