A synthetic hiring export we seeded passes the vendor's bias audit and fails the one NYC's Local Law 144 actually requires. The vendor's test checks race on its own and sex on its own, and both clear the legal 0.80 line at 0.8196 and 0.8744. Local Law 144 asks for race and sex together, and there the Black/Female group advanced at 0.65 of the White/Male rate of 1.00, an impact ratio of 0.6471. We planted that gap ourselves, in a seeded file of 1,040 synthetic candidates. No real employer, no real applicants. Six regulators look at that same hiring stack and ask six structurally different questions. NYC wants those combined ratios. Illinois bans zip-code proxies for race, while the EU wants training data that covers the map and leans on that same geography. No setup satisfies both. Clarion, the demo we built, runs one audit, fans it into six differently shaped deliverables, and puts nine of the 13 obligations in front of a human. Where two regimes want opposite things it returns CONFLICT in code, not on a model's judgment, plus a memo with the exposure on each side. If you screen candidates with automated tools in more than one jurisdiction, check the last audit your vendor sent: does it report race and sex together, or one at a time? If it is one at a time, it has not answered what New York City asks, let alone the other five. #AIHiring #AIGovernance #LocalLaw144 #HRCompliance