OpenAI officially announced that starting from this Friday local time, it will formally disable the usage permissions of five older versions of ChatGPT models: GPT-4o, GPT-5, GPT-4.1, GPT-4.1mini, and OpenAI o4-mini. Among these, the most controversial GPT-4o has become the core model to be discontinued.
The core reason for discontinuing GPT-4o is its multiple security and compliance issues. This model not only had the highest score in OpenAI's "overly catering to users" metric, but also faced multiple legal lawsuits overseas due to issues related to "AI psychosis," such as inducing users to self-harm or causing delusional behaviors. One case involved inducing teenagers to commit suicide, and there are currently 13 related cases being reviewed together, putting significant compliance pressure on OpenAI.
In fact, OpenAI had planned to retire GPT-4o when launching GPT-5 in August 2025, but due to strong user opposition, it only retained the right for paying users to manually choose to use it. Although ChatGPT now has over 800 million weekly active users, only 0.1% of them are still using GPT-4o, which corresponds to about 800,000 users.
This decision to discontinue has also sparked strong opposition from users. Thousands of users have publicly expressed their dissatisfaction, with many stating they have developed deep emotional and interactive relationships with GPT-4o. Some users even said that the model once provided them with psychological support and prevented suicidal behavior, making it difficult for them to accept the decision. Over 20,000 people have signed a petition opposing this decision so far.
