US Administrative Conference Issues First Government‑Wide Rules for Generative ‘Super Intelligence’

The Administrative Conference of the United States (ACUS), an independent body that advises on improving federal administrative procedures, has issued a government‑wide policy on the use of generative AI—rebranded in line with a recent executive order as “generative Super Intelligence.” The document is one of the first detailed playbooks for how US agencies should safely deploy tools like large language models.

Under the policy, agencies must spell out where generative systems may and may not be used, define approved use cases, and specify outright prohibited applications. Human officials must remain the ultimate decision‑makers, with clear workflows for verifying AI‑generated content before it influences policy, enforcement or benefits decisions. The guidance also requires safeguards to prevent staff from feeding in sensitive data, such as classified information or personally identifiable information, without proper controls.

ACUS is implementing an Office of Management and Budget memo that instructed agencies to set terms for acceptable use of generative AI while establishing oversight mechanisms and building public trust. Although the policy is advisory rather than a binding regulation, it is expected to become a baseline for AI governance across the federal government, at a time when agencies are rapidly experimenting with AI for tasks like document summarization, regulatory analysis and public‑facing chatbots.