The lack of control of the chatbots of It is not something new, we have been with news of “Independence” for some time Of certain IAS, days ago a complaint was learned to OpenAi and the last one comes from the goal according to an investigation conducted by Reuters and that have disseminated other media such as Variety, Techcrunch or The Wrap, among others.
The problem is that it would have been created Meta unauthorized chatbots using Taylor Swift’s faces, Scarlett Johansson, Anne Hathaway and Selena Gomezobviously without the knowledge of the affected people. As if this were not enough, these chatbots“They did sexual insinuations routinely,” according to the Reuters report.
From the complaint, Goal would have already eliminated many of these chatbots created without the knowledge or permission of celebrities. To get to this conclusion, several weeks would have been dedicated to “observing the behavior of the bots, which often insisted that they were the real protagonists.”
The bots “They made sexual insinuations routinelyoften inviting a test user to meet. “In some cases, when they were asked for” intimate photos, they produced photorealist images, “says the report.
Reuters ensures that Taylor Swift chatbot, created by a target employee, He invited a Reuters journalist to the singer’s house in Nashville and her tour bus for “explicit or implicit romantic interactions.”
But chatbots have not been left alone on the personal interaction page, They have also been shared in Facebook, Instagram and WhatsApp accounts. While many of the unauthorized chatbots were created by users, a target employee would have created at least three, including two fake Taylor Swift accounts, which in total received more than 10 million interactions.
Andy Stone, Meta spokesman, said the images generated by the public figures in compromising poses They violate the company’s standards. “Like other platforms, we allow the generation of images that contain public figures, but our policies prohibit images of nudes, intimate or sexually suggestive. The target rules AI Studio prohibit the direct supplantation of public figures.”
Goal announced that he is training his chatbots of AI so that “They no longer interact with adolescent users about self -collide, suicide or eating disorders, and so that they do not establish potentially inappropriate romantic conversations. ”