Promote secure-by-design AI applied sciences and functions: The plan says the US authorities “has a duty to make sure the AI methods it depends on — significantly for nationwide security functions — are protected in opposition to spurious or malicious inputs” and that “selling resilient and safe AI improvement and deployment needs to be a core exercise of the US authorities.” It recommends that DoD, in collaboration with NIST and ODNI, proceed to refine DoD’s accountable AI and generative AI frameworks, roadmaps, and toolkits. It additionally asks the ODNI, in session with DoD and CAISI, to publish an ordinary on AI assurance.
Promote mature federal capability for AI incident response: The plan asks NIST, together with CAISI, to associate with the AI and cybersecurity industries to make sure AI is included within the institution of requirements, response frameworks, greatest practices, and technical capabilities of incident response groups. It additional asks CISA to change its cybersecurity incident and vulnerability response playbooks to include concerns for AI methods and to incorporate necessities for CISOs to seek the advice of with chief AI officers, senior company officers for privateness, CAISI, and different officers as acceptable.
Assess nationwide security dangers: One other key provision asks for “American AI builders to allow the non-public sector to actively shield AI improvements from security dangers, together with malicious cyber actors, insider threats, and others.” It additional asks CAISI, in collaboration with nationwide security businesses, to “consider and assess potential security vulnerabilities and malign international affect arising from the usage of adversaries’ AI methods in essential infrastructure and elsewhere within the American financial system, together with the potential for backdoors and different malicious habits.”
