Paid user here — GPT-5 safety filters are ruining the product

The Impact of Enhanced Safety Filters on User Experience in Advanced AI Platforms

As a paying subscriber to ChatGPT, I have observed significant changes in the user experience with recent updates, particularly with the transition from GPT-4 to GPT-5. While advancements in AI safety are crucial, the implementation of increasingly aggressive safety filters appears to be hampering usability and user satisfaction.

Historically, GPT-4 provided a dynamic and engaging conversational experience, offering users creative and insightful responses across a variety of topics. However, the latest version, GPT-5, has introduced what many perceive as overly cautious filtering mechanisms. This has resulted in a more restrained AI, often characterized as “dry” and “lifeless,” with a tendency to issue moralizing or safety-related disclaimers even in benign contexts.

Users have reported that commonplace phrases, such as expressing tiredness or reluctance to attend university, now frequently trigger safety banners. Furthermore, messages that may hint at mental health struggles are subject to heightened scrutiny, often prompting warnings that can feel intrusive or misaligned with user intent. These measures, while well-intentioned, seem excessively aggressive, reducing the AI’s responsiveness and naturalness.

For many adult users who rely on ChatGPT for creative projects, professional assistance, or casual conversation, these safety filters can significantly diminish the value of the service. Instead of fostering engaging and meaningful interactions, the filters create a barrier that stifles spontaneity and authenticity.

There is a growing consensus among subscribers that they would benefit from opt-out options or adjustable safety settings, allowing users to tailor their experience according to personal comfort levels. Providing such flexibility would help preserve the core strengths of the platform—its intelligence, responsiveness, and creativity—while still maintaining safety standards.

In conclusion, while safety measures are a vital component of AI deployment, they should not come at the expense of user experience. Striking the right balance is essential to ensure that loyal users can continue to enjoy the full potential of advanced AI technology without unnecessary restrictions.