Governments should monitor large scale data centers and track AI incidents, and should require that AI developers of frontier models be subject to independent third party audits evaluating their information security and model safety.