AI Bill of Rights

Also known as: Blueprint for an AI Bill of Rights

framework · governance and compliance · organizing-schema

US Office of Science and Technology Policy's blueprint articulating five principles for automated systems affecting Americans' rights.

The AI Bill of Rights — formally the Blueprint for an AI Bill of Rights — is the non-binding policy guidance document published by the White House Office of Science and Technology Policy (OSTP) on October 4, 2022 articulating five principles to protect Americans from harms of automated systems. The Blueprint frames itself explicitly as a non-binding white paper rather than a regulation, policy, or executive order, and identifies five principles: Safe and Effective Systems; Algorithmic Discrimination Protections; Data Privacy; Notice and Explanation; and Human Alternatives, Consideration, and Fallback. Each principle is supported by an extensive accompanying Technical Companion document with implementation examples and case studies. The Blueprint substantially preceded President Biden's Executive Order 14110 of October 30, 2023 on Safe, Secure, and Trustworthy AI. President Trump's January 23, 2025 Executive Order 'Removing Barriers to American Leadership in Artificial Intelligence' revoked EO 14110, producing substantial uncertainty about the AI Bill of Rights's continuing federal-policy status, though the Blueprint as OSTP publication remains formally available.

Originators

White House Office of Science and Technology Policy (OSTP); Director Eric Lander (during early development), Alondra Nelson (Acting OSTP Director during October 2022 release); Technical Companion development through extensive civil-society, academic, and industry consultation including listening sessions and public input; intellectual antecedents in: civil-rights legal tradition; Universal Declaration of Human Rights; OECD AI Principles (2019); IEEE Ethically Aligned Design; substantial academic and civil-society work on algorithmic accountability including Cathy O'Neil's Weapons of Math Destruction (2016), Ruha Benjamin's Race After Technology (2019), Safiya Umoja Noble's Algorithms of Oppression (2018) high

Year / Decade

October 4, 2022 (OSTP publication); October 2023 (Biden EO 14110, related but distinct); January 2025 (Trump revocation of EO 14110, status complications) high

Primary sources

White House Office of Science and Technology Policy (2022). Blueprint for an AI Bill of Rights: Making Automated Systems Work for the American People, Biden, J. (2023). Executive Order 14110 on the Safe, Secure, and Trustworthy Development and Use of AI (October 30, 2023, subsequently revoked), Trump, D. (2025). Executive Order on Removing Barriers to American Leadership in AI (January 23, 2025), O'Neil, C. (2016). Weapons of Math Destruction (foundational civil-society antecedent) high

Band notes

Non-binding blueprint rather than enforceable regulation; superseded in regulatory effect by subsequent executive orders.

Core components

Five principles with accompanying detailed guidance: (1) Safe and Effective Systems: pre-deployment testing, risk identification and mitigation, ongoing monitoring
(2) Algorithmic Discrimination Protections: equity assessments, accessibility testing, disparity testing, organizational oversight
(3) Data Privacy: built-in protections, data agency, default privacy
consent and notice for sensitive data
protection from unchecked surveillance
(4) Notice and Explanation: clear documentation of system use, meaningful explanations of decisions, ongoing knowledge of impact
(5) Human Alternatives, Consideration, and Fallback: opt-out availability, access to remedy, fallback to human consideration
Technical Companion: detailed implementation guidance with case studies, examples of practices, and questions for consideration
Non-binding status: explicitly framed as 'not US Government policy' and not creating legal rights or obligations
Application scope: 'automated systems that have the potential to meaningfully impact the American public's rights, opportunities, or access to critical resources or services'
Relationship to subsequent Biden EO 14110 (October 2023, subsequently revoked): the EO operationalized some Blueprint principles through specific federal-agency directives, NIST AI Safety Institute establishment, and developer-reporting requirements for frontier models

Primary use case

Non-binding policy reference for federal-agency AI deployment and procurement (with substantial uncertainty about current status post-2025 EO changes); intellectual reference for state-level AI legislation (California Generative AI Training Data Transparency Act AB 2013, New York City Local Law 144 on AI in employment, Colorado AI Act 2024) drawing on Blueprint principles; civil-society and academic reference for algorithmic accountability advocacy; input to corporate responsible-AI policies in US-headquartered technology companies citing the Blueprint as foundation for internal practices; international reference framework comparison alongside EU AI Act (binding) and OECD AI Principles (intergovernmental soft law); academic reference in comparative AI regulation, algorithmic accountability, and US technology policy literature.

Common criticisms

The Blueprint's non-binding status has been substantially debated — civil-society groups including the Electronic Privacy Information Center, Algorithmic Justice League, and Center for Democracy and Technology argued during release that the absence of enforcement mechanisms limited the Blueprint's substantive impact while industry-aligned commentators argued non-binding guidance was the appropriate first step
the Trump January 2025 executive-order revocation of Biden EO 14110 and broader signals about AI policy direction have created substantial uncertainty about the Blueprint's continuing federal-policy status, with the document remaining formally available on OSTP website but its operational role unclear
the document's broad principles produce operationalization gaps — what 'algorithmic discrimination protections' specifically require in particular deployment contexts depends on implementation that the Blueprint structures only loosely
absence of specific scope definitions (which automated systems trigger which obligations) produces applicability ambiguity that has limited consistent implementation
the Blueprint's foundation in civil-rights framework has been criticized by some technical AI-safety researchers as insufficient for advanced-AI safety concerns (frontier-model capability risks, alignment concerns) that go beyond civil-rights-framework applicability
international interoperability with EU AI Act and OECD Principles is conceptual rather than formal
scholarly engagement (Frank Pasquale, Solon Barocas, danah boyd) has been productive but the Blueprint's policy impact has been argued to be modest relative to the binding-regulation alternative
the document's extensive Technical Companion provides substantial implementation guidance but the lack of legal-mandate infrastructure means practical adoption depends on voluntary uptake
criticism that the principles framework reproduces the limits of pure civil-rights law (individual remedy focus, insufficient structural-remedy power) without addressing systemic AI-deployment harms
reception in different communities (federal agencies, state legislatures, civil-society, industry) varies substantially.

Lineage

Siblings: NIST AI Risk Management Framework, EU AI Act, OECD AI Principles