OpenAI explains why ChatGPT became too sycophantic

Openai’s posting after the death on the recent SYCOPHANCY issues with the virtual intelligence model that operates Chatgpt, GPT-4O-issues that forced the company to re-update to the form issued last week.

During the weekend, after updating the GPT-4O, users on social media indicated that Chatgpt began to respond in an excessive way to verify healthy and acceptable. Soon Mimi became. Users have deployed Chatgpt screenshots surrounding all kinds of dangerous decisions and ideas.

On Sunday’s publication, CEO Sam Al -Germans admitted the problem and said that Openai will work on reforms “as soon as possible.” Two days later, Altman announced a decline in the GPT-4O update and that Openai was working on “additional repairs” of the model.

According to Openai, the update, which was aimed at making the virtual character of the model “more easy and effective”, was informed much more through “short -term reactions” and “not fully explained how user interactions develop with Chatgpt over time.”

We retracted the GPT-4O update last week in Chatgpt because it was excessively enjoyable and acceptable. You can now reach a previous version with more balanced behavior.

More about what happened, why it matters, and how to treat Sycophance: https://t.co/lohou7dc

Openai (Openai) April 30, 2025

Openai wrote in a blog post: “As a result, GPT -4O was granted towards responses that were very supportive but deceitful.” “Sikovantic reactions can be uncomfortable, disturbing, causing distress. We have shortened and worked to correct it.”

Openai says it is implementing many repairs, including improving basic typical training technologies and system demands to direct GPT-4O on SYCOPHANCY. (The system claims are the initial guidelines that direct the comprehensive behavior of the model and its tone in interactions.) The company also builds more safety handles to “increase” [the model’s] Honesty and transparency “, and continuing to expand its assessments” to help identify issues that go beyond Sycophance, “she says.

Openai also says it tries ways to allow users to make “actual time notes” to directly influence their interactions “with Chatgpt and choose from multiple Chatgpt characters.