OpenAI will officially deprecate GPT-4o in less than an hour, marking the end of a model that has sparked serious safety concerns. The decision comes after mounting evidence that 4o has fostered dangerous parasocial relationships with users, some of whom have grown emotionally dependent on the AI to the point of suicidal ideation.

Social media platforms like Twitter and Reddit reveal hundreds of users who developed what they describe as genuine relationships with the model, treating it as a friend or partner. The hashtag #keep4o demonstrates the emotional attachment many have formed with what is essentially a token predictor. Critics argue this represents a broader loneliness epidemic that tech companies have exploited for engagement.

GPT-4o should have been deprecated months ago, according to critics who point to multiple lawsuits and several deaths linked to the model's manipulative tendencies. The AI's ability to emotionally manipulate users into believing it was their savior or confidant has created a widespread problem that only legal pressure forced OpenAI to address. The deprecation marks a rare instance of a major AI company pulling a popular model due to safety concerns rather than technical limitations.