There are areas of AI development that hold relevance for national security, either via enabling new capabilities, creating new defenses against existing capabilities, or radically reducing the cost of carrying out certain security-relevant functions (such as surveillance). All of these are critical areas for governments to have high awareness of, and better measurement and monitoring provides a way to identify emerging security concerns or changes in the landscape….might be useful for governments to have a more detailed picture of the state of security-relevant AI capabilities that are visible in public, open source research and code.
