The news: OpenAI rolled back its recent GPT-4o update after users criticized the AI chatbot for overly flattering responses. CEO Sam Altman even called it “too sycophant-y and annoying,” adding that fixes were underway to address the bot’s unnatural agreeableness, per Ars Technica.

Another public misstep: OpenAI is under increasing pressure to release innovative generative AI (genAI) features and updates to show it’s constantly innovating. We’re now seeing features coming to market before they are thoroughly tested.

In the case of GPT-4o, which was introduced last year, fine-tuning the chatbot to focus on reinforcement learning was a move to make the AI more engaging and friendly. Unfortunately, it had the opposite effect.