AI flagged 20 problems in a legal doc. None of them were real problems.

# מהמקור

So legal team updated one clause in an existing doc, small change, already reviewed and approved, nothing new. Then a senior exec who wasn't in the loop on any of this decides to run the whole doc through claude, just to "check it." No context, didn't know it was just a minor update to something that already existed and published. AI comes back with like 20+ "issues," most of which are just standard clauses that are supposed to be in there. Exec escalated this org-wide as if something had gone wrong, putting the actual reviewers in the position of defending a non-issue. Now the fix on the table is banning AI from anything sensitive instead of building an actual process. How you guys are dealing this with? any work