The Trump administration directs U.S. government departments and agencies to pursue agreements that would test artificial intelligence models for cybersecurity risks, according to reports. In connection with an executive order signed by President Trump, officials are instructed to seek arrangements with relevant parties, with a focus on obtaining AI models for testing purposes. Multiple outlets report that the approach being discussed involves asking AI developers and companies to voluntarily submit their models rather than requiring mandatory submission. The cybersecurity testing is intended to evaluate how AI systems may be exploited or could fail under adversarial conditions and other security scenarios. The executive order outlines responsibilities for federal departments, officials, and agencies to secure the relevant agreements and carry out related efforts. While details of the testing process and scope are not specified in the brief summaries available from the outlets, the reports are consistent that the government is moving to establish a framework for assessing AI models’ cybersecurity characteristics through a voluntary submission mechanism. The administration’s actions are framed as part of broader efforts to improve the security of AI technologies used or developed in the United States.