Meta's new AI agent, Muse, can expose users' personal and business secrets when they share conversations through web links. Researchers reported the problem, and Meta has now suspended the sharing feature while it reviews the issue.
Muse is a general-purpose AI agent from Meta. Like many online services, it lets users share searches, conversations, images, and similar content with other people. The trouble is that this sharing can leak information users never meant to make public.
How Meta Muse Sharing Exposes Private Data
The share feature works the way it does on many platforms. A user clicks an option to create a unique URL, and anyone with that link can load a page hosting the shared content. To keep sensitive information away from the wrong people, these URLs usually contain a long sequence of text that cannot be easily guessed.
But with Muse, that protection is not enough. According to the reporting, a single query is all that is required to expose the shared content. That means a person does not need to guess a long link or break any security system. One simple request can pull out information that was supposed to stay private.
The risk covers both personal and business material. Conversations people share casually could contain private details, and business users could expose confidential information without realizing it.
Meta Suspends Muse Sharing Feature After Researchers Flag Risk
When Ars asked Meta about the behavior, the company said it is suspending the sharing feature. Meta added that it is reviewing the feature and considering making changes to it.
That response matters for two reasons. First, it confirms the problem is real enough that Meta chose to switch the feature off rather than defend it. Second, it shows the company is still deciding how to fix the issue, so there is no clear timeline for when sharing will return or what it will look like.
- Muse is Meta's new general-purpose AI agent.
- The share feature creates unique URLs that host searches, conversations, and images.
- Researchers found that a single query can expose the shared content.
- Meta has suspended the sharing feature while it reviews the behavior.
- Meta says it is considering making changes to the feature.
Our Take: A Privacy Warning for Every AI Chat User
In our view, this is a serious warning that goes beyond one product. AI agents are being pushed into everyday use fast, and share buttons are treated as a small convenience. But when an AI tool holds your conversations, a share link is not a small thing. It can carry your private life and your company's secrets.
To put it plainly, the fact that a single query can expose shared content is a basic design failure, not a minor bug. Meta did the right thing by suspending the feature quickly. But users should not have to rely on researchers finding these holes first.
What this means for readers is simple. Treat every share link from an AI tool as public until proven otherwise. Do not share conversations that contain passwords, client details, financial data, or anything you would not post openly. Until Meta finishes its review and explains what changed, the safest move is to keep Muse conversations to yourself.
Sources & References
- Ars Technica — Original reporting on the Muse sharing flaw and Meta's response.
