If you would not be okay with that, what level of consequence would be acceptable for the output from this tool?
I’d want detectors to be as accurate as possible, false positives of 1 in 10000 seems like a good starting point. I believe their results have been independently tested.
And as a separate matter, any tool for evaluating students should be applied fairly, safely, and with adequate human review and due process.
You need good tools and good oversight.
Plagiarism and cheating sucks for everyone. Worth solving.