If an agent under a company’s control commits a crime, the company has committed that crime. If we hold companies directly responsible for the actions of their agents, we may end up seeing companies being more secure with their testing and model development.