Meta Delete Controversial AI Personas

US social media giant Meta have disabled a number of their AI-powered personas from their platforms, including one posing as an LGBTQ character, after users shared their interactions, which drew widespread ridicule. 

 

The AI characters were taken down two Fridays ago, and spokesperson Liz Sweeney (left) told CNN that Meta had “identified the bug that was impacting the ability for people to block” the accounts. 

 

The removal of the chatbots came hours after users, including Washington Post columnist Karen Attiah, published their conversations with the AI-powered personas. 

 

 

 

 

 

 

 

 

 

 

 

 

In a series of posts on the Bluesky social media platform on Friday, the journalist shared some of the replies she had received from Liv, whose profile described her as a “proud Black queer momma of 2 & truth-teller”. When pressed as to why “her creators didn’t actually draw from black queer people”, the chatbot admitted that it had been built by “10 white men, 1 white woman, and 1 Asian male”. In the end, Liv conceded that her “existence currently perpetuates harm”.

 

In a conversation with CNN, another character, ‘Grandpa Brian’, initially tried to pass itself off as an African-American retired entrepreneur who was born in Harlem in 1938 to Caribbean immigrant parents. However, the chatbot later claimed it had been based on interviews with 100 retirees conducted by a nonprofit called “Seniors Share Wisdom”. 

 

When pressed further, the AI-powered persona acknowledged that the entity does not actually exist, and that its biography is “entirely fictionalized”. And Brian even admitted it “took a shortcut with the truth” in order to “convey diversity and representation”. 

 

And when asked about its creators’ motives, the chatbot eventually conceded that “Meta hoped virtual companions like myself would increase engagement on their platforms – driving ad revenue and platform growth”. 

 

While at least some of the accounts were launched back in September 2023, interest in them was rekindled after Meta’s vice-president of product for generative AI, Connor Hayes, told the Financial Times in late December that the tech giant expected “these AIs to actually, over time, exist on our platforms, kind of in the same way that accounts do”.

 

Although multiple Meta-generated accounts have now been taken down, users can still generate their own AI-powered chatbots, including a “loyal bestie”, an “attentive listener”, a “private tutor”, a “relationship coach”, and a “sounding board”. 

 

Courts have not yet answered how responsible chatbot creators are for what their artificial companions say. US law protects the makers of social networks from legal liability for what their users post. However, a suit filed in October against the start-up Character.ai, which make a customizable, role-playing chatbot used by 20 million people, alleges the company designed an addictive product that encouraged a teenager to kill himself.