The legislation mandates the Secretary of Defense to manage risks associated with the military use of artificial intelligence (AI).
A structured, indexed ledger of all uses of covered AI systems by the Department of Defense must be created within three years.
The ledger must be regularly maintained and updated for accuracy.
A comprehensive risk assessment process for evaluating each unique implementation of covered AI systems must be established within three years.
The risk assessment process must address dependability, cybersecurity, privacy, bias, escalation bias, deployment span, and risk of civilian harm.
Annual reevaluations of each covered system in the ledger are required, along with assessments triggered by updates or new procurements.
The Secretary must annotate the ledger when covered systems are shared or exported to foreign entities, noting any additional risks.
Progress reports on the implementation of the Act must be submitted to Congress annually for three years.
An annual submission to Congress must include the ledger, risk assessment findings, and annotations regarding exports, with an emphasis on public availability of unclassified information.
The legislation expresses a sense of Congress supporting transparency and responsible policies regarding military AI use.
Existing privacy, safety, or security protections are not diminished by this legislation.
Definitions for “artificial intelligence” and “covered system” are provided, specifying the types of systems included.
