That's always been a possibility, even before the advent of LLMs. It was just harder to hide before. I think you should ask: what good is the project if you don't release it other than self-satisfaction? Sometimes just making something without releasing it is good enough (that's me in many cases). But since this is research in frontier field, I think it would be worthwhile publishing. Have you considered a less restrictive license like Apache 2.0 or MIT? Another possibility, depending on the nature of the project, is ensure the output of the project is not GPL while the project it self can GPL (like the GCC compiler). Or to put it another way which would bother you more: Spending years on research and never releasing it? Or some a-hole using your research without proper attribution?
That's always been a possibility, even before the advent of LLMs. It was just harder to hide before. I think you should ask: what good is the project if you don't release it other than self-satisfaction? Sometimes just making something without releasing it is good enough (that's me in many cases). But since this is research in frontier field, I think it would be worthwhile publishing. Have you considered a less restrictive license like Apache 2.0 or MIT? Another possibility, depending on the nature of the project, is ensure the output of the project is not GPL while the project it self can GPL (like the GCC compiler). Or to put it another way which would bother you more: Spending years on research and never releasing it? Or some a-hole using your research without proper attribution?
The license never protected your code. Lawyers did, if you could afford them. It's no different today, it's just harder to prosecute.
If your worry about people copying your code is preventing you from sharing, then open source and free software might not be for you.