States are racing to pass AI rules—over 1,500 bills in 45 states so far this year—but the courts lack the data judges need to decide if those laws run afoul of the dormant Commerce Clause. Under that doctrine, state rules can’t unfairly burden out-of-state actors, discriminate against nonresidents or extend beyond borders. The hardest test is “Pike balancing” from 1970’s Pike v. Bruce Church, which tells courts to compare a law’s interstate costs against its local benefits. Judges routinely face Pike without any systematic evidence on either side—no dollar estimates, no standardized metrics, nothing to gauge whether a $3 million compliance cost outweighs a cybersecurity gain. That evidentiary gap matters most to startups and smaller AI firms. Large platforms can absorb a maze of state requirements; emerging players can’t. Last week xAI sued over Colorado’s AI Act, claiming it violates the Commerce Clause. To avoid random outcomes, the piece argues, lawmakers and regulators must build records on costs, benefits and alternative approaches before courts weigh in. The White House Executive Order on AI already tasks Commerce with flagging burdensome state laws and sets up a Justice Department task force to spot dormant-Commerce violations. Those steps could generate useful data—but they won’t be enough. To close the gap, Congress and agencies should mandate impact analyses whenever states propose AI rules, using consistent metrics for economic and non-economic effects. They could require expert testimony or create technical reports summarizing how a bill might affect cross-border data flows, developer costs or consumer safety. On the judicial side, courts need clear guidelines: checklists, model findings, maybe a specialized benchbook on quantifying regulatory impacts. If judges get both better evidence and methods to interpret it, dormant-Commerce challenges to state AI laws will land on firmer ground—and smaller AI companies won’t be left guessing which state rules they can survive.