Researchers develop a framework for guarding against the potential harms of AI that also overcomes the inflexible nature of government regulation.