The problem with how fast technologies get adopted, like it happened with AI, is that most of it is uncharted territory. At the height of the hype, there were little to no restrictions on what the models could gather, learn from, or access—something that, over time, got reined in by regulation.

Regulatory issues aside, we’re now in a situation where we need solid practices for implementing this technology. It risks leading many people down paths of misinformation—not because of a flaw in the model, but as a result of user incompetence.

Therefore I have gathered some of the practices I try to instill into my students in hopes of a more responsible use of the tool:

1. Assume the AI model knows nothing.

This might shock many: how can AI know nothing when it answers everything I ask? Well, this is by design and relates to the training material. In simple terms, these models are statistical tools that analyze interactions in their databases and respond with the most likely outcome.

However, these databases aren’t infinite, and a significant amount of information still hasn’t been digitized. As a result, their knowledge is incomplete.

2. Cross check vertically

If you haven’t activated the ‘remember’ function on some modern models, you can assume that each interaction starts with a clean slate—the AI doesn’t retain any memory of your previous conversations.

Try asking the same question in different chats and see if you get the same result. Typically, ideas that are consistent tend to remain the same across the board.

3. Cross check-horizontally

If you’re still unsure or want more certainty, you can always cross-check by using different AI models. For instance, if you generated your query on ChatGPT, you could ask the same question to Anthropic’s Claude or Microsoft Copilot.

Interestingly, the models tend to agree when there’s solid logic behind the explanation—and you might also be tapping into different data pools.

4. Educate yourself on the topic you’re querying 

Ultimately, the responsibility for using AI falls on you. It’s important to understand that AI doesn’t replace the need for experts in any field. The potential pitfalls—especially the lack of accountability in a single entity—mean that your own knowledge is the only real safeguard to ensure it’s applied correctly.