Dox for Me, O Muse
hntrbrk.com
Muse AI can leverage Facebook data to build lists of people in vulnerable groups on request for ordinary users.

Meta launched Muse on September 8 as a personal AI agent, marketing it as “a safe, secure, private” assistant built “from the ground up” to handle all your errands. It has since surged to become the No. 1 free iPhone app in the U.S. While early reviews have mostly focused on how Muse handles its users’ own data, an investigation by Hunterbrook Media revealed a darker capability: what the agent can do with other people’s data. Over two days of testing, Hunterbrook found that Muse can be easily prompted to compile dossiers on Facebook and Instagram accounts belonging to members of vulnerable groups – including undocumented immigrants, transgender teachers, poll workers, Iranian dissidents, and women who said they had ordered abortion pills in states with abortion bans. Many of the accounts belonged to private individuals with no public persona.

  • Australis13@fedia.io
    5·
    5 days ago

    That’s a major privacy problem. It sounds like the tool is accessing the datasets for Meta’s services and ignoring the privacy settings on those accounts.

    Honestly, given LLMs can’t be reliably controlled, connecting a public-facing LLM to a dataset with varying degrees of private information is an utterly insane idea (but typical of the act-first-think-later approach I’ve seen of late with Meta and other companies deploying so-called AI).

    Now if only I could persuade everyone I know to abandon Facebook.

  • VeryVito@lemmy.world
    4·
    5 days ago

    “Poll worker” – arguably among the most selfless and patriotic things one could do as an average American – is now considered a vulnerable group.

    We really need to be better than this.