If the ethics crisis in AI is to be more than just a chorus of concerns, the move to address the issue must be structural, with governments setting enforceable limits on surveillance, bias, and data extraction.