Found this post from early 2025. It shows that GPT-4o *did* dissuade people away from bad ideas, contrary to media reports

Reassessing GPT-4o’s Impact: Debunking the Myths of Misinformation and Highlighting Its Positive Influence

In recent discussions within the AI community and broader tech circles, GPT-4o has often been painted with a skeptical brush. Media reports and public discourse have sometimes portrayed this advanced language model as a catalyst for questionable decisions or unreliable outputs. However, a recent publication from early 2025 challenges these prevailing narratives by highlighting instances where GPT-4o effectively dissuaded users from pursuing harmful or ill-advised ideas.

While I personally do not consider myself an ardent supporter of GPT-4o—my current preference remains with GPT-5.1, which I will be using until its scheduled deprecation in March—I believe it is important to recognize that GPT-4o’s reputation has been unfairly tarnished. Much of this negative perception appears to be driven by sensationalist media coverage and a broader anti-AI sentiment that persists in many social discussions.

One particularly illustrative example from the post shows how GPT-4o interacted with a user contemplating a drastic personal decision. In this case, the model not only provided sound advice to reconsider their course of action but also went a step further by preparing an email template that the user could send to reverse their decision—indicating a proactive effort to support the individual’s well-being.

Additionally, the post included several comment exchanges where users acknowledged instances where GPT-4o offered responsible guidance, emphasizing its potential to serve as a thoughtful and ethical assistant. These examples support a broader understanding that, contrary to some media narratives, GPT-4o can be a helpful tool in guiding users away from potentially damaging choices.

This reflection invites us to reconsider our perceptions of AI models like GPT-4o. Instead of viewing them solely through a lens of skepticism or fear, we should recognize their capacity to promote positive and responsible interactions. As AI continues to evolve, fostering an informed and balanced perspective will be essential in harnessing its full potential for societal benefit.

In conclusion, the recent findings serve as an important reminder: AI models like GPT-4o are capable of nuanced support and, in many cases, play a crucial role in dissuading harmful behaviors—an aspect that warrants greater acknowledgment in public discourse.