US National Archives clarifies when government AI use counts as an official record

The US National Archives and Records Administration (NARA) has issued new guidance clarifying how federal record‑keeping rules apply when agencies use AI tools. The key message: simply using an AI system does not automatically create a federal record. Instead, as with other technologies, what matters is whether the AI-generated content is actually used to conduct government business, inform decisions, or shape policy.

NARA offers concrete examples. If a staffer uses a generative AI chatbot for preliminary research and none of the output is incorporated into a memo or report, the chat history would generally fall outside federal records requirements. By contrast, when an agency relies on AI‑generated analysis or text in drafting policy documents or making official decisions, that material may need to be preserved as part of the record.

The clarification comes as civil servants increasingly experiment with external AI services, raising questions about transparency, privacy, and information governance. Agencies are now expected to align their AI usage policies with existing records rules—deciding which prompts and outputs must be retained, how to document AI’s role in decision‑making, and how to safeguard sensitive data fed into these systems.