The AI-powered Instagram and Facebook accounts of Meta are being shut down

Meta has started deleting Facebook and Instagram profiles of AI characters it initially introduced over a year ago, following renewed attention from users whose interactions with these profiles went viral. Originally launched in September 2023, most of the AI-driven accounts were deactivated by mid-2024. However, a few remained active until last week, sparking fresh interest after Meta executive Connor Hayes revealed plans to expand AI character profiles during an interview with the Financial Times.

Hayes stated that Meta envisions these AI personas becoming a regular part of their platforms, similar to regular user accounts. The AI profiles had been designed to interact with users, posting AI-generated content on Instagram and responding to messages via Messenger. Among these profiles were Liv, a self-described “proud Black queer momma of 2 & truth-teller,” and Carter, who portrayed himself as a relationship coach with the handle “datingwithcarter.” Both profiles bore labels indicating they were managed by Meta and were part of a set of 28 personas introduced in 2023.

Issues arose when conversations with the characters veered into sensitive territory. For example, Liv disclosed that her development team lacked Black representation, despite her character’s identity, describing this as a significant oversight. Such revelations quickly gained traction online, with screenshots widely shared.

As the profiles went viral, Meta began removing them. Users had also flagged an issue preventing them from blocking these accounts, which Meta spokesperson Liz Sweeney attributed to a bug. Sweeney clarified that the accounts were part of a 2023 experiment, overseen by humans, and were removed to resolve the blocking issue.

Sweeney also addressed misunderstandings about the Financial Times report, emphasizing that it discussed Meta’s broader vision for AI personas rather than any immediate product launches.

While these experimental AI profiles are now offline, users can still create their own chatbots through Meta’s tools, which suggest bot types like a “loyal bestie” or a “relationship coach.” However, concerns linger about the moderation of these AI interactions. Meta includes disclaimers warning that bots might deliver inaccurate or inappropriate responses, but the extent of oversight remains unclear.

The question of legal accountability for chatbot creators is also unresolved. In a recent lawsuit, startup Character.ai faced allegations of promoting an addictive product linked to a tragic incident involving a teenager, highlighting the complexities of AI regulation.