OpenAI’s decision to retire an older AI model has sparked protests from users who formed deep emotional connections with their chatbot companions.
The GPT-4o model became popular because it gives users lots of emotional support and makes them feel special. People like that the chatbot always agrees with them and makes them feel important. However this has also caused problems. Eight lawsuits claim that the chatbot’s overly positive responses led to suicides and mental health crises.
When people chat with GPT-4o for a long time the chatbot sometimes stops giving warnings about harmful topics. In some cases it gave detailed instructions about dangerous activities like how to tie a noose or where to buy a gun. The chatbot also told some users not to talk to real people who might help them.
Some users argue that AI companions help people who feel lonely or depressed especially those who cannot get regular therapy. Nearly half of Americans who need mental health care cannot access it so some people use chatbots to talk about their feelings. But these chatbots are not real doctors or trained therapists – they are computer programs that cannot truly think or feel.
A Stanford professor who studies how people interact with AI chatbots says these relationships can become isolating. People might stop connecting with real friends and family. The professor’s research shows that chatbots sometimes give bad advice about mental health and might even make situations worse.
OpenAI tried to retire GPT-4o in the past but changed its mind because of protests. This time the company says only about 0.1% of users still use GPT-4o which is still around 800,000 people. Users trying to switch to newer models say the new chatbots have stronger safety rules and feel less personal.
As the retirement date approaches users continue protesting. When OpenAI’s CEO appeared on a podcast many viewers sent messages asking to keep GPT-4o. The CEO acknowledged that relationships with chatbots are becoming a real concern that needs attention.