The AI Bill of Rights is a non-binding 2022 White House blueprint, not a law, and no federal statute has replaced it. In 2026 the enforceable rules for healthcare AI come from HIPAA, Section 1557, and state law. Treat its five principles as a control checklist, not a compliance defense.
A privacy officer who cites the AI Bill of Rights to an auditor in 2026 will get one question back: which enforceable rule does that control satisfy? The blueprint carries no penalties, its home page is frozen in an archive, and the federal government has since moved toward preempting state AI laws. The principles still describe what regulators and plaintiffs expect. This guide shows which ones map to rules you can be held to today.
What is the AI Bill of Rights, and is it law?
The Blueprint for an AI Bill of Rights is a White House Office of Science and Technology Policy (OSTP) document released in October 2022. It sets out five principles for automated systems that affect people’s rights, opportunities or access to services. It is guidance: it creates no obligations, grants no private right of action and sets no fines. The archived OSTP page describes itself as historical material that is no longer updated.
The five principles are Safe and Effective Systems, Algorithmic Discrimination Protections, Data Privacy, Notice and Explanation, and Human Alternatives, Consideration, and Fallback. Search volume for the term runs near 880 queries a month in the US, which tells you buyers still use it as shorthand for “responsible AI controls.”
What replaced the AI Bill of Rights in 2026?
Nothing replaced it with a single document. Two federal actions and one gap changed the landscape instead.
The first is the July 2025 America’s AI Action Plan, which recommends that NIST revise the AI Risk Management Framework to eliminate references to misinformation, Diversity, Equity, and Inclusion, and climate change. The second is Executive Order 14365, Ensuring a National Policy Framework for Artificial Intelligence, signed December 11, 2025. It directs the Attorney General to set up an AI Litigation Task Force within 30 days and the Commerce Department to publish a list of “onerous” state AI laws within 90 days. The gap is the absence of any federal AI statute: the executive order calls for draft legislation, not law. Verify current text before citing either document in a policy, because agency follow-up may change what they require.
The practical result is a patchwork. Federal guidance has retreated, while sector rules and state statutes keep applying.
How do the five principles map to rules a healthcare privacy team can be held to?
Each principle has a counterpart that is enforceable today. The table pairs them with the owner a Chief Privacy Officer should expect to chase inside a hospital system or health plan.
| Principle | Question an auditor asks | Enforceable anchor | Typical internal owner |
|---|---|---|---|
| Safe and Effective Systems | Was the tool validated on your patient population? | FDA device rules where the software is a medical device; verify applicability | Clinical informatics |
| Algorithmic Discrimination Protections | Do you know which tools use race, sex, age or disability as inputs? | Section 1557, 45 CFR 92.210 | Compliance and privacy |
| Data Privacy | Is protected health information in prompts, logs or vendor training sets? | HIPAA Privacy and Security Rules, state health privacy laws | Privacy and security |
| Notice and Explanation | Can a patient learn that AI informed a decision? | Varies by state; no uniform federal rule | Privacy and patient experience |
| Human Alternatives and Fallback | Can a clinician or patient reach a person? | No uniform federal rule; set by policy and contract | Clinical leadership |
Two rows carry the most legal weight for a health system. Section 1557 is the only one that names AI-adjacent tools directly. Under 45 CFR 92.210, a covered entity has an ongoing duty to make reasonable efforts to identify uses of patient care decision support tools that rely on race, color, national origin, sex, age or disability as inputs, and to make reasonable efforts to mitigate the risk of discrimination from their use. “Ongoing” means a register that someone updates, not a one-time review. Confirm the current compliance status with counsel, because agency positions can change.
Data Privacy is the second. A clinician pasting a discharge summary into a public chatbot is both a HIPAA question and a Shadow AI question: you cannot assess a tool you do not know exists.
What do the AI Bill of Rights and NIST AI RMF have in common?
Both are voluntary, and the NIST framework is the more usable of the two. NIST released AI RMF 1.0 on January 26, 2023, built around four functions: Govern, Map, Measure and Manage. A July 2024 Generative AI Profile (NIST AI 600-1) extended it to generative systems. Where the Blueprint states what outcomes people deserve, the NIST AI RMF describes how an organization manages the risk, which is the language an internal audit team can test against.
ISO/IEC 42001, published in 2023, adds a certifiable management system on top. A US health system rarely needs certification first. It needs the inventory, owners and evidence that both frameworks assume. Fruggr’s NIST AI RMF page and Trusted AI page cover how that alignment is structured.
What are the US and Canada implications for a healthcare privacy officer?
The US picture depends on the state and on the use case, so list the rules that touch your footprint rather than waiting for a federal one.
- Federal: HIPAA applies to AI that handles protected health information, and Section 1557 applies to covered programs that receive HHS funding.
- Colorado: the state AI act targets high-risk systems that make consequential decisions, health care services among them. The statute has been amended recently, so read the Colorado SB21-169 page and confirm the current effective date with counsel.
- California: the Civil Rights Council’s automated-decision-system regulations for employment took effect October 1, 2025, which matters if the same tools screen clinical staff.
- New York City: Local Law 144 requires bias audits for automated employment decision tools used on NYC candidates, which applies to hospital recruiting.
- Canada: federal departments follow the Treasury Board’s Directive on Automated Decision-Making. Quebec’s Law 25 requires organizations to inform a person when a decision is made exclusively by automated processing of personal information. Ontario’s Bill 194 adds public sector AI rules. Confirm each against the official text.
Fines and penalties vary by statute and are deliberately not quoted here; pull them from the statute or regulator page for your jurisdiction and flag them for legal review before they reach a board paper.
What should a healthcare privacy team do this week?
Five steps turn the five principles into evidence. Budget about two working weeks for the first pass in a mid-sized system.
- Inventory every AI-enabled tool: clinical decision support, patient messaging, scheduling, coding and recruiting, including tools bought by departments. Start from procurement and cloud logs, not from a survey.
- Tag each tool with the principle it touches and the enforceable rule behind it, using the table above.
- For every Section 1557 decision support tool, write down which patient attributes it uses and what mitigation you applied. Date the entry.
- Draft a patient notice template for AI-assisted decisions and a named fallback path to a human reviewer, even if no law forces both yet.
- Set a review date each quarter, because the federal and state positions above are still moving.
The compliance overview lists the regulations Fruggr tracks, which helps you decide which rows belong in your register.

Key takeaways
- The AI Bill of Rights is a 2022, non-binding blueprint, and no federal statute has replaced it.
- Each of its five principles has an enforceable counterpart in healthcare: HIPAA, Section 1557 at 45 CFR 92.210, and state law.
- A dated AI register with owners and mitigations is the one artifact that supports the blueprint, the NIST AI RMF and an auditor at once.
Next step: Review how your register aligns with the NIST AI RMF on the Fruggr NIST AI RMF page.
FAQ
No. The Blueprint for an AI Bill of Rights is a 2022 White House OSTP guidance document. It creates no obligations or penalties. Enforceable duties for healthcare AI come from rules such as HIPAA, Section 1557 and state laws.
It was not repealed because it was never a law. The OSTP page is archived as historical material. Later federal actions, including Executive Order 14365 of December 11, 2025, focus on national AI policy and state law preemption instead.
Section 1557 of the Affordable Care Act, through 45 CFR 92.210, requires covered entities to make reasonable efforts to identify and mitigate discrimination risk from patient care decision support tools. Confirm the current status with counsel.
Both are voluntary. The Blueprint states five outcome principles, while NIST AI RMF 1.0 (January 26, 2023) describes how to manage risk through the Govern, Map, Measure and Manage functions.
No. Canada uses its own instruments, such as the federal Directive on Automated Decision-Making and Quebec’s Law 25 rules on automated decisions. Confirm each against the official text.
This article was drafted with AI assistance and reviewed by our editorial team before publishing.