Lawyers cannot trust AI-generated legal work because citations are fabricated and there is no safe, verified way to use it

Summary

Litigators are adopting consumer-grade generative AI to draft substantive filings (e.g., uploading pleadings, discovery, and depositions to draft a response to a summary judgment motion) while knowing that made-up citations and hallucinated authorities are a live professional-liability risk. The frustration is acute and openly acknowledged: lawyers say nobody — not CLE panels, not experts — will address the real question of whether using ordinary consumer AI on privileged case files violates professional duties, and many admit they hide how much they rely on it. The workaround today is manual: re-checking every citation, quotation, and factual assertion against the source file, which erases much of the time saved. No existing tool both generates case-grounded drafts and anchors every sentence to a verifiable, source-pinpointed citation inside the uploaded file, nor handles confidentiality posture transparently. This is unsolved, severe (direct exposure to sanctions and malpractice), and exactly the gap the case-intelligence workspace must close: AI answers and drafts must be checked against the file before anyone relies on them.