AI rules and guidance by sector compared: banking, health care, federal agencies and Colorado
| Rule or guidance | Sector | Binding or not | Duty | Date |
|---|---|---|---|---|
| Revised interagency Guidance on Model Risk Management (Fed SR 26-2; OCC Bulletin 2026-13), replacing SR 11-7 and SR 21-81 | SectorBanking: issued by the Federal Reserve, OCC and FDIC; expected to be most relevant to banking organizations with over $30 billion in total assets1 | Binding or notGuidance. It sets no enforceable standards or prescriptive requirements, and non-compliance will not result in supervisory criticism.2 | DutyA risk-based approach tailored to each banking organization's model risk profile, size and complexity: model development and use, including testing; validation and monitoring; governance and controls; and considerations for vendor and other third-party products, including their validation2 | DateApril 17, 20261 |
| Generative and agentic AI models at banks2 | SectorBanking2 | Binding or notNot covered: the 2026 model risk guidance places generative and agentic AI models outside its scope2 | DutyNone set by the guidance. The agencies plan a request for information on model risk management that considers banks' use of AI, including generative AI, agentic AI and AI-based models.2 | DateRequest for information planned 'in the near future', as stated on April 17, 20262 |
| 45 CFR 92.210 (ACA section 1557 rule): nondiscrimination in the use of patient care decision support tools3 | SectorHealth care: covered entities, meaning recipients of HHS financial assistance, HHS itself and entities set up under title I of the ACA. The tools covered are any automated or non-automated tool used to support clinical decision-making.4 | Binding or notBinding regulation3 | DutyNo discrimination on race, color, national origin, sex, age or disability through such tools in its health programs or activities; an ongoing duty to make reasonable efforts to identify uses of tools whose input variables or factors measure those characteristics; and reasonable efforts to mitigate the discrimination risk of each tool identified3 | DatePart 92 effective July 5, 2024; covered entities must comply with the identification and mitigation duties in 92.210(b) and (c) within 300 days of July 5, 20244 |
| ONC decision support interventions certification criterion, 45 CFR 170.315(b)(11), as revised by the final rule of October 28, 2026 (91 FR 67916)5 | SectorHealth IT: developers certifying Health IT Modules in the ONC Health IT Certification Program5 | Binding or notCertification criterion, kept in revised form. The source attribute ('model card') requirements in (b)(11)(iv) and (v) and the intervention risk management requirements for predictive interventions in (b)(11)(vi) are removed.5 | DutyRetained: letting a limited set of identified users select evidence-based and predictive decision support interventions, and the electronic feedback tools for evidence-based interventions; predictive interventions are limited to data in (b)(11)(iii)(A) plus the Clinical Notes and Assessment and Plan of Treatment data classes. The expired clinical decision support criterion in (a)(9) is removed.5 | DatePublished October 28, 2026; effective December 28, 20265 |
| OMB Memorandum M-25-21, Accelerating Federal Use of AI through Innovation, Governance, and Public Trust (replaces M-24-10)6 | SectorFederal agencies; some requirements apply only to CFO Act agencies and some not to the Intelligence Community. Uses presumed high-impact include, in health care, medical device functions, patient diagnosis, risk assessment or treatment, and health-insurance cost control and underwriting; the high-impact definition also covers effects on access to credit and insurance.6 | Binding or notBinding on agencies: they must follow the implementation instructions and requirements in its appendix6 | DutyMinimum practices for each high-impact use: pre-deployment testing, an AI impact assessment, ongoing monitoring, training and oversight of operators, and timely human review and appeal for people affected; a use that does not meet them must be safely discontinued6 | DateIssued April 3, 2025; agencies must document implementation of the minimum practices within 365 days of issuance6 |
| OMB Memorandum M-25-22, Driving Efficient Acquisition of Artificial Intelligence in Government (replaces M-24-18)7 | SectorFederal agencies buying AI; not AI acquired as a component of a national security system7 | Binding or notBinding on agencies for covered contracts7 | DutyUpdate acquisition procedures within 270 days so acquired AI conforms to M-25-21; when practicable, disclose in solicitations whether a planned AI use meets the high-impact threshold; protect against vendor lock-in7 | DateIssued April 3, 2025; applies to contracts from solicitations issued 180 days or more after issuance, and to options exercised after that date7 |
| Colorado SB 26-189, Automated Decision-Making Technology (repeals and reenacts SB 24-205)8 | SectorCross-sector. For the sectors in this table, it reaches decisions on access to, eligibility for or compensation related to financial or lending services, insurance and health-care services, as well as education, employment, housing and essential government services and public benefits8 | Binding or notState law. A violation is a deceptive trade practice under the Colorado Consumer Protection Act, which the Attorney General enforces; no new private right of action. Entities the act specifies are exempt to the extent they comply with other legal obligations.8 | DutyDeployers: notice at the point of interaction; a plain-language description of the technology's role within 30 days after an adverse consequential decision; on request, meaningful human review and reconsideration, and access to and correction of the personal data used. Developers: technical documentation for deployers and notice of material updates. Both: compliance records for at least 3 years.8 | DateGovernor signed May 14, 2026. From January 1, 2027, developers owe the documentation; by the same date the Attorney General must adopt rules on post-adverse-outcome disclosures. Before any action begun before January 1, 2030, the Attorney General gives a 60-day notice and opportunity to cure.8 |
License: CC BY 4.0. Underlying government texts are public domain or under their own terms. Attribute "Moraph Insights" with a link to the page.
Method
We read each rule or guidance in its official source: the Federal Reserve's SR 26-2 letter and the OCC's Bulletin 2026-13, the eCFR text of 45 CFR part 92 as of 2026-11-12, the Federal Register text of ONC's October 28, 2026 final rule on govinfo, the OMB memoranda PDFs, and the Colorado General Assembly's bill page with its summary as enacted. Each cell was checked on 2026-11-14 against the source it cites. Dates are copied as the sources give them. The broader AI governance frameworks (NIST AI RMF, EU AI Act) are compared in the AI governance table. This table explains the rules; it is not legal advice.
Notes
- The 2026 model risk guidance defines a model as a complex quantitative method that applies statistical, economic or financial theories to produce quantitative estimates; simple arithmetic and deterministic rule-based software are excluded.
- The OCC bulletin rescinds the Comptroller's Handbook booklet on model risk management and OCC Bulletins 1997-24, 2011-12 and 2021-19.
- ONC's October 28, 2026 rule says comments on proposals it did not finalize remain under review.
- ONC's October 28, 2026 rule encourages users of certified health IT to ask developers for information on predictive tools and reminds them of their duty under 45 CFR 92.210 to identify potential bias and avoid discrimination.
Sources
- Board of Governors of the Federal Reserve System. SR 26-2, Revised Guidance on Model Risk Management (April 17, 2026). Checked .
- Office of the Comptroller of the Currency. OCC Bulletin 2026-13, Model Risk Management: Revised Guidance (April 17, 2026). Checked .
- U.S. Department of Health and Human Services (eCFR). 45 CFR 92.210, Nondiscrimination in the use of patient care decision support tools. Checked .
- U.S. Department of Health and Human Services (eCFR). 45 CFR part 92, subpart A (sections 92.1 and 92.4: dates and definitions). Checked .
- Office of the National Coordinator for Health Information Technology, HHS. Health Data, Technology, and Interoperability: ONC Deregulatory Actions To Unleash Prosperity, final rule (91 FR 67916, October 28, 2026), sections I, III.A.1.a and III.A.2.e and amendatory instruction 11. Checked .
- Executive Office of the President, Office of Management and Budget. OMB Memorandum M-25-21, Accelerating Federal Use of AI through Innovation, Governance, and Public Trust (April 3, 2025). Checked .
- Executive Office of the President, Office of Management and Budget. OMB Memorandum M-25-22, Driving Efficient Acquisition of Artificial Intelligence in Government (April 3, 2025). Checked .
- Colorado General Assembly. SB26-189, Automated Decision-Making Technology (bill summary as enacted and bill history). Checked .
How we source and check every fact: Editorial standards · Corrections
Last reviewed .
Spotted an error? Tell us and we will check it and correct the table where needed.