NGA proposes federal AI oversight while preserving state authority
The National Governors Association’s Nov. 22, 2026, announcement sets out principles for AI governance at both state and federal levels. The proposal assigns national-scale risks to federal oversight while retaining a role for states in protecting people from AI-related harms.
Key takeaways · 4
- 01
Use the proposal to distinguish federal oversight of frontier-model risks from state accountability for harms affecting people.
- 02
The NGA recommends independent expert evaluations and reporting security breaches and significant safety incidents to federal authorities and affected states.
- 03
For consequential automated decisions, the principles call for human oversight and ways for people to challenge adverse outcomes.
- 04
The NGA says any federal preemption of state AI laws should be approached cautiously and include input from a representative cross-section of states.
A proposed division of responsibility
The NGA says the principles are intended to build a shared national approach, promote consistency among state frameworks and inform the incoming Congress.[1] Its proposed division gives the federal government responsibility for national and international AI risks, while preserving state authority over AI within states’ responsibilities.[2] The distinction matters for organizations operating across jurisdictions: the document seeks coordination, but does not propose removing state roles.[2] The principles were led by Governors Wes Moore, Mike Braun, Josh Stein and Spencer Cox.[1]
Federal rules for advanced models
The NGA recommends a risk-based federal framework for the safe and responsible development of frontier models.[2] The proposed standards would apply to the most advanced general-purpose models whose risks and reach are national in scale.[2] It also recommends evaluations or audits by independent experts.[2] Developers, under the proposal, would report security breaches and significant safety incidents to federal authorities and affected states.[2] The principles additionally call for national standards addressing AI-related threats to critical infrastructure, including financial systems and cybersecurity.[2]
State protections remain central
The NGA recommends that states remain a primary accountability mechanism for AI harms affecting people as AI is deployed.[2] It says states should preserve consumer and worker protections, anti-discrimination enforcement, and responses to fraud and impersonation.[2] The principles also call for safeguards against AI-related online harms to children, including risks involving addictive design, chatbots and companion applications.[2] For automated systems used in housing, employment, lending, insurance and essential services, the NGA recommends human oversight, attention to unlawful discrimination and ways to challenge adverse decisions.[2]
Coordination and open questions
The NGA says requirements should use technically informed definitions and proportionate safeguards that are regularly reviewed as AI capabilities evolve.[2] It calls for reducing unnecessary duplication and conflicting requirements where possible, while encouraging competition and avoiding rules that favor larger companies or unnecessarily impede smaller entrants.[2] The association says federal preemption of state AI laws should be approached cautiously and include meaningful input from a representative cross-section of states.[2] Its working group plans to engage governors’ offices in all states and territories on common definitions, standards, best practices and multistate cooperation.[2]
The principles offer organizations a way to anticipate how federal model-level oversight could coexist with state-level safeguards for people affected by AI decisions. Teams can use the proposed division of responsibility to identify where they may need separate review, incident-reporting and decision-challenge processes, while treating the document as recommendations rather than assuming a specific legal requirement.
Why it matters
AI training for your team
kju.ai is built for teams: daily sessions, team analytics, and rollout support for your organization.
How this developed
22 November 2026
NGA proposes federal AI oversight while preserving state authority
