Most law firms have an AI policy. Almost none have tested whether it actually works. The Real-World Case File closes that gap with 14 real, cited court and disciplinary rulings pulled from the Charlotin AI Hallucination Cases Database, covering everything from the original Mata v. Avianca sanction to a 2026 filing where Sullivan & Cromwell, a firm with a written AI policy, mandatory training, and a “verify everything” rule, still submitted 40+ fabricated citations.

Each case is built the same way: what happened, the penalty (from a symbolic $1 CAD to a $110,000 sanction to an indefinite license suspension), and a “Why This Case Is in the File” note tying it to one of the Ten Hidden Risks named in the companion report. Two cases show courts declining to sanction, so the file reads as evidence, not scare tactics.

This isn’t a legal-research summary, it’s a talking-points tool. Each entry is built to be dropped into a partner meeting, a training deck, or a client conversation, where a named case with a dollar figure and a judge attached lands harder than a general warning ever will.