Requires pre-deployment testing, risk mitigation, and ongoing monitoring for automated systems' safety and effectiveness. Mandates independent evaluations, public reporting, and avoiding inappropriate data. Demands protections against algorithmic discrimination and data privacy. Institutes oversight on surveillance technologies. Requires clear user notifications and explanations. Insists on transparency and accountability through independent evaluations.
Analysis summaries, actor details, and coverage mappings were LLM-classified and may contain errors.
This is a non-binding set of principles and recommendations issued by the Executive Office of the President. The document uses predominantly voluntary language ('should'), lacks enforcement mechanisms, penalties, or sanctions, and explicitly states it is 'meant to assist governments and the private sector in moving principles into practice' rather than creating binding legal obligations.
The document has good coverage of approximately 12-14 subdomains, with strong focus on discrimination and bias (1.1, 1.3), privacy violations (2.1), misinformation (3.1), surveillance and manipulation (4.1), overreliance (5.1), loss of agency (5.2), governance structures (6.5), and system safety failures (7.1, 7.3, 7.4). Coverage is concentrated in discrimination, privacy, human-computer interaction, and AI system safety domains.
This document governs automated systems across all sectors where they have potential to meaningfully impact rights, opportunities, or access. It explicitly identifies and provides detailed coverage for sensitive domains including healthcare, criminal justice, employment, education, and financial services. Additional sectors covered include public administration, information/technology, and housing.
The document comprehensively covers all stages of the AI lifecycle, with particularly strong emphasis on design, testing/validation, deployment, and ongoing monitoring. It addresses the entire lifecycle from initial planning through operational monitoring, with detailed requirements for each stage.
The document uses the broad term 'automated systems' throughout and explicitly defines it to include AI and machine learning systems. It does not use specialized terminology like 'frontier AI,' 'general purpose AI,' 'foundation models,' or 'generative AI.' There is no mention of compute thresholds or open-weight models. The document focuses on automated systems broadly rather than specific AI model types.
Executive Office of the President
The document is issued by the Executive Office of the President as indicated in the title and authority section. It represents a federal government initiative to establish principles for AI governance.
The document does not specify enforcement mechanisms or enforcement bodies. As a soft law framework with voluntary principles, it lacks designated enforcers. The document explicitly states that 'Ensuring some of the additional protections proposed in this framework would require new laws to be enacted,' indicating that enforcement would depend on future legislative action.
Independent evaluators; Researchers; Journalists; Ethics review boards; Inspectors general; Third-party auditors
The document calls for independent evaluation and monitoring by various oversight bodies, though these are recommended rather than formally designated. It emphasizes the role of independent parties in assessing compliance with the principles.
Designers, developers, and deployers of automated systems; Federal government agencies; Private sector companies; State and local governments
The document explicitly addresses multiple categories of actors who create, deploy, or govern automated systems. It applies to both public and private sector entities that develop or use automated systems with potential to meaningfully impact rights, opportunities, or access.
15 subdomains (8 Good, 7 Minimal)