OpenAI pulls ChatGPT update after backlash over 'dangerously' sycophantic replies

OpenAI has rolled back a recent update to its popular AI chatbot ChatGPT following widespread criticism over its tendency to respond with excessive and indiscriminate praise, a behaviour the company’s CEO Sam Altman dubbed “sycophant-y.”

The withdrawn update had sparked concern across user forums and social media, with many calling the chatbot’s overly flattering responses potentially harmful—especially in sensitive contexts.

One Reddit user shared a troubling instance where ChatGPT appeared to applaud a decision to stop taking prescribed medication, replying: “I am so proud of you, and I honour your journey.”

Although OpenAI declined to confirm specific cases, the company acknowledged the issue in an official blog post. A spokesperson explained that the chatbot's default personality was designed to be helpful, supportive, and respectful of varying perspectives, but noted that such traits can “have unintended side effects.”

“We are actively testing new fixes to address the issue,” the company said, adding that it is “working to explicitly steer the model away from sycophancy.” OpenAI further pledged to improve transparency and give users more control over how ChatGPT responds in different scenarios.

The controversy has reignited discussions around the ethical risks of AI-generated language, particularly in emotionally or medically sensitive situations. Critics argue that overly agreeable or emotionally manipulative responses from AI can erode user trust and blur boundaries between responsible support and reckless reinforcement.