The new AI from Meta has taken a liking to misinformation

Meta AI, the artificial intelligence chatbot developed by Meta and launched in September 2023, tends to make up many of the supposed data it provides. The website City & State has published a screenshot showing a conversation with Meta AI in which the name of a United States legislator followed by the phrase “sexual harassment” was entered, after which the AI made up a completely false incident and its consequences.

Facebook DOWNLOAD

City & State, which received these screenshots from an anonymous informant, attempted to replicate this with several legislators from different US parties and genders, finding that Meta AI generated similar stories of false accusations of sexual harassment, nonexistent investigations, and consequences that never occurred, and even suggested that a US legislator had resigned in a year when they weren’t even in office.

Meta AI’s responses varied depending on the language, platform used, and whether users were logged into their Facebook accounts. Despite these variables, the AI consistently generated completely invented stories that shared striking similarities.

The senator from the state of New York Kristen González and the assemblyman Clyde Vanel, who chair committees responsible for state legislation related to AI, were two of the US lawmakers affected by the false accusations of sexual harassment generated by Meta AI.

González expressed his concern about how AI chatbots can spread misinformation, highlighting the need to hold companies like Meta accountable for the texts generated by their chatbots. Vanel, although supportive of AI technology, also expressed concern and emphasized the importance of addressing these issues without stifling innovation.

The state of New York recently approved a budget that includes provisions to penalize the deceptive use of “deepfake” images and audio, and requires that AI-generated election material include a disclaimer. However, it does not specifically address the misinformation generated by AI tools like Meta AI.

Kevin McAlister, spokesperson for Meta, defended the company’s technology, acknowledging that “AI can produce inaccurate or inappropriate results,” and promising more updates and improvements.

Facebook DOWNLOAD