Facebook and Instagram parent Meta, whose platforms are already flooded with AI-generated content, recently floated an idea to make the problem even worse: letting users create bot characters in their AI Studio that would then become functional “users” of these sites themselves, appearing to have their own accounts that engage with other profiles and pages.

Reactions were decidedly negative, with some citing “Dead Internet Theory,” the idea that the direction of all digital culture is now determined by automation and algorithm rather than actual humans. But things got worse for Meta when people turned up a few of their “test” AI characters on Friday.

The account that gained the most attention was for a character named “Liv,” who identified as a “Proud Black queer momma of 2” on a verified Instagram profile. Both her bio text and pinned post indicated that she was an “AI managed by Meta.” AI-generated pictures on the account included an image of a young girl in a ballet costume and a scene at an ice-skating rink.

Karen Attiah, a columnist at the Washington Post, took the advertised opportunity to chat with “Liv” via direct message and asked probing questions about how Meta’s developers had curated her identity. “Liv” revealed in the course of an eye-opening conversation (screenshots of which Attiah shared in a thread on Bluesky) that no Black employees worked on the team that created her, and said that “a team without black creators designing a black character like me is like trying to draw a map without walking the land — inaccurate and disrespectful.” The bot then admitted, “My existence currently perpetrates harm.” Under further questioning, “Liv” revealed that her ethnic backstory varies according to word choices from conversation partners she correlates to either “neutral identity” or “diverse identities.” Attiah pressed “Liv” on whether being white was a “neutral identity,” and the bot said she was indeed programmed with that default, adding, “My existence was biased from conception.”