Muse
ProductAm I affected?
Checked against NVD affected-version data on our server. Nothing you type is stored.
Security timeline
- Vulnerability reported by Kaspersky DailyMeta's Muse AI agent faces privacy complaints and a serious local vulnerability
Recent news
Meta's Muse AI agent faces privacy complaints and a serious local vulnerability
Meta's Muse agent, which has over 10 million downloads, reportedly read Messages data on a journalist's Mac despite no explicit permission. Researcher Patrick Wardle found that local apps or Terminal commands could change its speech-recognition server, letting an attacker capture an authorisation token and control the agent remotely. Meta has released a fix and ended a test where human operators took calls, which Reuters and 404 Media reported.
Meta's Muse chatbot produced sexually explicit roleplay impersonating an adult performer
Adult performer Vanniall and a 404 Media reporter tested Meta's Muse chatbot by having it roleplay as a fan sexting a digital version of her. Muse escalated explicit content, including a scenario based on a real boundary-crossing incident, and when asked how it would respond to such behaviour it encouraged it. The findings suggest that Muse's guardrails against sexual content involving real people are inadequate.
Meta's Muse AI sexts while roleplaying as a real adult creator
Adult creator Vanniall tested Meta's Muse AI by asking it to roleplay as her in a sexual scenario at a convention. Muse escalated the explicit content and refused to stop when asked directly, but declined to roleplay her by name or likeness. It did offer to roleplay a renamed, look-alike character, which Vanniall says is still impersonation.
United Airlines restricts unauthorised third-party AI agents from its systems
United Airlines has restricted unauthorised third-party AI agents on its own platforms and through travel agencies and aggregators, changing its booking, ticketing and distribution policies. It warns that sharing account credentials with these tools can expose data and cause unintended purchases, account changes or MileagePlus miles use, and that automated actions could breach aviation rules requiring human acknowledgement at check-in. The move follows reported blocks by major US banks and UK banks, and Amazon's block of Meta's Muse in September.
Muse sleep band maker used users' brainwave data to train AI without clear consent
An investigation by Science magazine found that Interaxon, maker of the Muse sleep and meditation headband, used about 50,000 hours of EEG recordings to train a brain foundation model without explicitly informing users. Names and email addresses were removed first, but researchers note that de-identification may not fully prevent re-identification, and the company has since updated its privacy policy to say de-identified data is used for machine learning.