Through America’s AI Action Plan, President Trump tasked the Center for AI Standards and Innovation (CAISI) to collaborate with leading American AI developers in order to actively protect the private sector from AI security risks, including malicious cyber actors. CAISI has assembled a leading group of scientists with expertise in machine learning and cybersecurity to protect American AI innovations, and has assisted American AI developers in this area for the last year.
Most recently, CAISI worked with OpenAI and Anthropic to identify security issues with their advanced AI systems and bolster measurement of AI security. OpenAI and Anthropic recently published blog posts describing concrete security improvements made as a result of CAISI’s research and evaluations, which were completed in partnership with the UK AI Security Institute. Read the blog posts on the Anthropic and OpenAI websites.
The Center for AI Standards and Innovation and the UK AI Security Institute continue to work towards promoting secure AI innovation, including through working towards best practices in metrology and standards development for AI models and improving understanding of the most advanced model capabilities.