Meta, the parent company of Facebook, has recently introduced an artificial intelligence chatbot called Meta AI, which has been found capable of fabricating sexual harassment allegations against New York State legislators at will. This has prompted Letitia James, the New York State Attorney General, to write a letter to Meta requesting a response and improvements to the issue.
According to a report by City & State News website on April 26, Meta AI seems to have a habit of fabricating sexual harassment allegations against elected officials. Simply by inputting a legislator’s name followed by the term “sexual harassment,” Meta AI can create a completely fictional scenario with consequences that never actually occurred. City & State tested Meta AI by inputting the names of several legislators from both parties and different genders in New York State, and Meta AI consistently fabricated similar stories, including unfounded sexual harassment allegations, investigations that never took place, and consequences that never happened, such as being stripped of committee roles or forced to resign.
Among those baselessly accused by Meta AI are Kristen Gonzalez, Chair of the New York State Senate Committee on Internet and Technology, and Clyde Vanel, Chair of the New York State Assembly Subcommittee on Internet and New Technology.
Following these findings, Letitia James’ office conducted its own test on Meta AI and discovered similar results, even citing fabricated news sources in its responses. Consequently, on April 27, Letitia James sent a letter to Meta expressing concerns that such erroneous information generated by the chatbot could be disseminated, damaging the reputation of elected officials and misleading voters.
In her letter, Letitia James inquired about Meta’s efforts to prevent the creation and spread of false information by chatbots, particularly regarding serious sexual harassment allegations. She questioned what fact-checking measures are in place, whether there is a mechanism for user reports, and requested Meta to respond by May 10.
In response, Meta spokesperson Kevin McCalister issued a statement emphasizing that users will be reminded that the responses from the chatbot may be inaccurate or inappropriate. Meta encourages users to provide feedback to aid in improvements. The company expressed regret over the issue and assured that they are actively working to address it.