I'm in the process of publishing an article in the Engineering field. The manuscript got a preliminary acceptance, and should be published after few modifications. However, based on one comment I got, I need to evaluate my model's performance based on certain input data. I don't have such data, but such data has been published before in the literature. Is it OK to use the date published by other authors just to evaluate my model? What am I going to do is to digitize the published figures and use the values. I'm going to cite the papers, but I'm not going to add the figures or the datapoints to my manuscript.
Yeah, it is fine. If anything, it is how the literature and scientific enterprise is supposed to work. I would just be clear about attribution and that you haven't discovered the relationship (from first paper) but instead used existing info to validate your method.
You can almost certainly use the data that you find. Copyright restrictions only prevent you from republishing it. But if the originators make it available, they do so so that others can take advantage of it (as well as making verification of their own work possible).
Of course, the original authors may grant a non-restrictive license to the data in any case, in which case you are bound by that.
The same is true if you read something in a copyrighted book. You can use that information for any purpose, but you cannot republish it. But for your work all you really need to do is to cite it properly to avoid any charge of plagiarism.
Given the final comments in your question, you should be fine.
If the data is publicly available, yes, by all means, but don't forget to cite the source of the data (and any associated papers that introduce or describe it).