To answer this question, one can first refer to the chart shown below, derived from this UNESCO publication.

Here’s the essence of the chart: ChatGPT responses may contain errors and false or inaccurate information. If this is not a concern, then it is certainly safe to use the model; otherwise, one should ask whether one has the necessary skills to verify if ChatGPT’s output is correct and accurate or not. If you lack these skills, then it is not safe to use ChatGPT. However, if you do possess these skills, as a final step, you need to ask yourself whether you can – and want to – take full responsibility, from every perspective, for any errors or inaccuracies that may go unnoticed. If the answer is no, then, once again, it is not safe to use ChatGPT. If the answer is yes, then you can use the model’s output after carefully verifying it. The same guidelines apply to other systems, such as Google’s Bard, but also more generally. In his Thrustworthy Generative AI course, Professor Jules White (Vanderbilt University) invites us to think about this problem in terms of risk: what consequences could using AI-generated results have? Could there be serious repercussions for people? Economic or reputational harm? Other significant negative effects? Furthermore, is it easy or difficult to verify the obtained answers? If using the output of an AI can have significant consequences, verifying the answers is difficult, and you lack the skills to do so, then it is certainly advisable not to use AI or to use it in a way that reduces risks. How? Professor White provides these two examples: instead of asking ChatGPT if it is safe to take drug X (high risk), you can ask what questions it would be appropriate to ask a doctor, i.e. an expert, to evaluate the safety of taking X given your health conditions (lower risk, possibly affecting your reputation). Furthermore, instead of asking ChatGPT to extract/summarize critical information from a certain document (high risk), you can ask it to identify where in the document that information can be found (lower risk).
Using AI correctly requires not fewer skills but more skills, despite the countless charlatans who present AI as a sort of genie in a bottle that provides the solution to this or that problem without any effort on your part. A story that has been heard a million times, and just as many times proven false, if not dangerous.