Nearly three out of four American teenagers have already used “IA companions”, according to the NGO Common Sense Media, who worries the risks that these artificial intelligence assistants (IA) oriented on personal relations can represent for minors.
• Read also: The boss of Nvidia salutes the “catalyst” role of Chinese AI
• Read also: AI seeks the means to reduce its energy needs
• Read also: Meta will invest hundreds of billions of additional dollars in IA infrastructure
Some “72% of teenagers used IA companions at least once” and “more than half use these platforms several times a month,” said the organization in a study carried out in the spring in the United States and published on Wednesday.
The IA companions are defined by the association as “friends or digital characters with whom we can discuss in writing or oral at any time”, to have “personal and significant conversations”, services provided in particular by Character.
The questionnaire presented to participants aged 13 to 17 years acknowledges that the star platforms of the generative AI, such as Chatgpt (Openai) or Claude (Anthropic) can also be used in this way, but specifies that it is not the tools for homework or generation of images.
A third of the teenagers interviewed says they have exchanges with the IA companions who fall under social relations, as training to have certain conversations, receive moral support, or even have friendly or romantic interactions.
Almost half consider these assistants AI above all as “tools” or “computer programs”.
Common Sense Media, an NGO that classifies movies, video games and applications according to age to help parents, and campaigns for online security, is concerned about the potential effects of the use of these “friends, confidents and even virtual therapists”.
“The dangers they represent for young users are real, serious and well documented”, assures the association, citing the example, of the “suicide of a 14 -year -old teenager who had developed an emotional attachment to an IA companion”.
The NGO highlights the trend of these tools generally appear to agree with users, instead of encouraging them to think, which is problematic according to it at an age when they develop their ability to think critically.
“Several popular IA companions platforms, including Character.ai, Nomi and Relima (…) present unacceptable risks for users under the age of 18,” concludes Common Sense Media.
“They can easily generate answers ranging from content of sexual nature and offensive stereotypes to dangerous” advice “which, if they were followed, could have serious, even deadly, in real life”.
She therefore recommends not to let minors use these tools.
These three companies did not immediately respond to AFP requests.
