Accurate, reliable, legally usable summation of a large corpus is worth billions, I am sure you will get assistance with specific technical challenges you may encounter along the path to that goal, but a general purpose, fully working, open source solution that has been debugged and is ready for monetisation and field trails is very likely to be kept close to the developers chest, I know of a couple and both have received 9 figure investment already.
If you eliminate the possibility of hallucination, you can get accurate and reliable. But, “legally usable”? In the current environment, do you really think courts are going to admit AI generated summations as “legally usable”? I mean, I haven’t a clue, but seeing what has already happened in the copyright courts, it’s not looking good.
We are some time away from an AI lawyer, but paralegal? That I think is not so far off.