White House issues 10 principles for US agencies to consider when making rules for the use of AI in the private sector but says they should avoid “overreach”
James Vincent / The Verge :
Context & Ripple Effects
This January 2020 memo is the opening move of a six-year arc in US AI policy: ten nonbinding principles telling agencies to weigh costs against benefits and avoid regulatory overreach when touching private-sector AI. At the time it reads as a deregulatory posture — guidance, not rules.
What came after shows the posture hardening. Eleven months later the White House issued an executive order on building public trust in federal AI development, then in 2023 the Office of Management and Budget moved from principles to draft rules requiring agencies to assess AI harms in health care, law enforcement, and housing. By 2026 the same office is calling on Congress to preempt state AI laws and impose age-gating on models.
First-order effects
- US agencies drafting any rule that touches private-sector AI are told to apply cost-benefit analysis and scientific standards, raising the evidentiary bar before regulation can proceed.
- Private-sector AI developers get an explicit federal signal of restraint: no new compliance obligations attach to these principles, and the administration frames heavy-handed rules as a risk to innovation.
Second-order effects
- The voluntary framing creates a vacuum that later documents fill: the December 2020 executive order on trustworthy federal AI and OMB's 2023 harm-assessment drafts show each successive layer converting soft guidance into concrete agency duties.
- Agencies in high-stakes domains — health care, law enforcement, housing — become the testing ground where abstract principles get operationalized into assessment requirements, setting templates other regulators copy.
Third-order effects
- If the pattern holds, US AI governance ratchets from nonbinding principles to a prescriptive federal framework — culminating in the 2026 push for Congress to preempt state AI laws and gate model access — with the White House, not Congress, defining the terms at each step.
- The through-line also exposes the limits of executive-branch guidance: because each administration's principles can be replaced rather than repealed, durable AI regulation ultimately depends on legislation, which is exactly what the 2026 framework asks Congress to provide.
The trend: US AI regulation is drifting from anti-overreach voluntary principles toward an increasingly prescriptive federal framework, with the White House steering the pace at every stage.