Intellectual property (IP) is anything that you conceptualize or create: whether it is written, played, sung, viewed, or synthesized in a lab. Different countries, universities, and companies have legislation regarding intellectual property.
In academia, the IP usually belongs to the researcher, with approval of the university, and thanks to the 1980 Bayh-Dole Act, universities can patent and sell IP. Since you are an employee, and the university owns the building, arranges for the electricity bill to be paid, manages your grant funds so you don’t run into accounting problems, and technically owns the equipment that you buy with grant funds, they have a stake in your work even though they don’t outright own it. This means you are free to publish anything that is true and verifiable, and both governmental funding agencies and college policy generally dictate that you have to honestly publish your results no matter the outcome. The best you can do is not publish a project, but you can’t keep anyone else from replicating your work and then publishing it.
In fact, that is where a lot of confidentiality concerns surrounding IP come into play. For example, if you are working on something, and another group publishes a similar project first, you’ve been scooped and your work is no longer novel. In the rare cases that someone stole your work, and you can prove it, the university will back you on IP rights. While competition can be fierce in many fields or work settings, outright theft of data is not common despite the widespread fear of it.
Confidentiality is a large part of academic work, but in most cases is a temporary part. You might not be able to publish a commercial lab’s proprietary procedure, but you can and are encouraged to make publicly available all your methods and all your raw data. Once you publish it in some way, your name is now tied to it, and you will have a citation credit. The availability of raw and processed data is hugely important to most fields today. Not only does it allow you to analyze your work in reference to someone else’s results, but it provides the opportunity for large-scale data mining. There are quite a few studies that downloaded and reanalyzed a large number of similar data sets in order to better compare them and gain new insights into trends, as well as drug-prediction studies that use DNA or protein sequences from other projects as models against their drug to test its ability to bind and work effectively before trying it out in the lab.
Personally, I try to be transparent about what I work on at the local scale, and a bit vague on social media until the experiment is concluded and results published. For my thesis, I opted to wait six months after the University of Vermont had accepted it to make it open access. Print copies could be ordered, but it would not yet appear online. This was because several chapters were manuscripts which had not yet been published, and once it appears in print online that is considered “published” and some journals would not want to publish it again in an identical format. I tend to be very descriptive in how I present my bioinformatic workflows, because when I was learning the trade, a methods section that played things close to the chest made it very difficult for me to understand how others had analyzed their data. Once published, though, I think it’s very important to disseminate and publicize your work as much as possible; it doesn’t do the scientific community or the general public any good if they never see it.