NARA clarifies that AI tool use alone does not create federal records
The National Archives and Records Administration issued new guidance stating that merely using AI platforms does not automatically generate federal records.
A memo dated August 21 from the National Archives and Records Administration outlines when AI-related material must be preserved as federal records. The guidance makes clear that the use of AI systems does not, by itself, produce a record; instead, the material’s role in decision-making, official tasks, circulation, or system integration is decisive. Under the Federal Records Act, agencies must document functions and policies, but mere creation or receipt of AI output is insufficient.
Examples illustrate that an employee’s preliminary AI research that is not incorporated into a white paper would likely fall outside record-keeping requirements. Agencies may keep audit trails of AI interactions, but such logs become records only if deliberately used for investigations or other official purposes. While commercial AI services such as Gemini and ChatGPT are typically not federal records, each agency remains responsible for its own determination, and NARA recommends formal AI policies developed with legal and IT input.
Why it matters
Federal agencies need clear rules to decide which AI outputs must be archived for accountability.
In this story
