The parents of a 16 -year -old who took his life after months of interacting with Chatgpt sued Openai and his owner, Sam Altman, arguing that E andL Chatbot, under the GPT-4o model, did not apply security measures despite recognizing the suicidal intentions of the young man.
According to Adam Raine’s parents, Chatgpt came to advise methods to take his life and became his confidant about his suicidal ideas.
Given this situation, the OpenAI company has recognized that its artificial intelligence chatbot (AI) Chatgpt commits failures in “sensitive” cases and promised changes after receiving a demand for the responsibility of its technology in the suicide of a minor in the US, EFE reports.
Openai published a blog post entitled “Help people when they need it most” And, although he did not refer to the demand, he indicated that ChatgPT is trained to recommend to those users to “express suicidal intention” the contact of professional assistance organizations.
Nevertheless, He acknowledged that despite having that and other security mechanisms for when he detects that “someone is vulnerable and the risk may be”, the chatgpt systems “fall short” And the chatbot “did not behave as it should in sensitive situations.”
The company explained that security mechanisms work better in short exchanges and can fail in long interactions that “degrade” AI training, so it works specifically that it takes measures if it detects that “suicidal intention” in multiple conversations.
Chatgpt, said the company, also recognizes when the user is under 18 to apply protection measures, but will now include parental controls so that adolescents responsible know how they use that technology, and explore the possibility of connecting them directly with an emergency contact.
Among other things, the company said that its “mitigation” systems have focused on cases of self-harm but now other “emotional anguish” will also cover, and the GPT-5 model will be updated to be able to “descale” situations of that type “connecting the person with reality.”
It also studies “connecting people with certified therapists before they are in an acute crisis”, which would imply “a network of graduated professionals that people can call directly through Chatgpt,” but considered that this possibility will take time.