OpenAI Releases 'National Security Principles' to Define Boundaries for AI Use in Government and Defense

OpenAI has released the 'National Security Principles', which clarify the boundaries and usage norms for collaboration with governments in the areas of national security and law enforcement. These principles apply to existing and future collaborations with agencies such as the U.S. Department of Defense, reiterating the prohibition of using technology for large-scale domestic surveillance, commanding autonomous weapon systems, and high-risk automated decision-making. OpenAI has established 'trusted access' collaborations in cybersecurity and biosecurity with the U.S. and its allies, working with multiple national cybersecurity agencies through the Daybreak initiative, and has made the biosecurity model GPT-Rosalind available to certain government agencies for public health and bio-defense tasks. OpenAI supports legislative measures to impose external constraints on military applications of high-risk AI.