The READ AI Models Act directs the National Institute of Standards and Technology (NIST) to run a pilot program to create a structured template and technical guidance for documenting artificial intelligence models and associated data. The work is intended to apply across public- and private-sector models, be informed by external stakeholders, and be published for public use.
This is a technical, standards-focused bill: it charges NIST with producing a flexible documentation template, accompanying metrics and benchmarks, and a process for stakeholder input, then reporting to Congress on the pilot’s effectiveness and making the resulting resources publicly available. For organizations that build, buy, or oversee AI, the bill aims to reduce information gaps that complicate procurement, auditing, and risk assessment.
At a Glance
What It Does
The bill directs NIST to initiate a pilot to design a structured documentation template and detailed technical guidelines for AI models and datasets, informed by industry and government stakeholders. NIST must publish the resources and report on the pilot’s effectiveness to Congress.
Who It Affects
Model developers and vendors, federal agencies that use or procure AI, private-sector purchasers, auditors and compliance teams, standards bodies, and researchers will all be potential users of the template and guidance. The resources are intended for voluntary adoption across sectors.
Why It Matters
Standardized documentation reduces information asymmetry in procurement and oversight, creates common inputs for risk assessment and auditing, and can accelerate adoption of interoperable best practices across jurisdictions and industries.
More articles like this one.
A weekly email with all the latest developments on this topic.
What This Bill Actually Does
The bill authorizes NIST to start a pilot program to develop a structured template and technical guidance that will accompany AI models and the datasets used with them. NIST is to work with other federal agencies and outside parties to design resources that help users understand what a model is, how it was created, and what limits or guarantees accompany its use.
The statute frames this work as applied technical guidance rather than new regulatory obligations.
NIST’s deliverable is twofold: a structured template for documenting models and a set of detailed technical guidelines to explain how to complete the template and how to measure or benchmark the items documented. The law explicitly contemplates a single template that is modular so organizations can adopt the portions that fit a particular use case or sector.
The guidance is expected to help third-party assessors, procurement officers, and internal compliance teams evaluate models more consistently.The statute also prescribes a consultative development process: NIST must collaborate with federal agencies and a broad set of external stakeholders (private companies, universities, nonprofits, and international standards organizations). That collaborative work is meant to inform both the content of the template and the technical metrics that underpin it.
Importantly, the bill ties the pilot to available appropriations and frames NIST’s work as a publicly shared resource to support transparency and interoperability across buyers and regulators.By providing a shared starting point for documenting model provenance, intended use, and known limitations, the resources NIST produces are intended to reduce friction in contracting, make audits and compliance reviews more efficient, and create a foundation that others—industry or regulators—can reference when specifying documentation expectations.
The Five Things You Need to Know
The bill gives NIST a non-exhaustive list of model metadata examples the template may include, such as model name, developer identity, developer incorporation location, release date, knowledge cutoff for training data, languages supported, and the model’s terms of service.
The statute requires the documentation template be modular so organizations can adopt sections or categories tailored to sector-specific needs, intended audience, or desired characteristics.
NIST’s technical guidelines must offer relevant metrics or benchmarks for each component of the template where applicable and are to incorporate voluntary consensus-based technical standards and industry best practices.
NIST must publish a draft of the template and technical guidelines in the Federal Register and allow at least 60 days for public comment before finalizing the materials.
Within 12 months of establishing the pilot, NIST must report to the House Science Committee and the Senate Commerce Committee assessing the pilot’s effectiveness and, if effective, include a plan for permanent implementation, and it must make the final template and guidance publicly available on NIST’s website.
Section-by-Section Breakdown
Every bill we cover gets an analysis of its key sections.
Short title
This section names the Act the "Resources for Evaluating and Documenting AI Models" or the "READ AI Models Act." It has no substantive requirements but frames the statute’s purpose for subsequent provisions and reporting.
Pilot program to develop template and guidance
This paragraph directs the NIST Director to initiate a pilot program to create a structured documentation template and accompanying technical guidelines for AI models and associated data. Practically, the pilot structure gives NIST discretion over scope and participants while signaling that the output is intended for broad public and private-sector use; the provision is conditioned on available appropriations, so implementation depends on funding decisions.
Template content and technical guidance requirements
These clauses set out the types of information the template may capture (examples provided in the text) and require that the template be modular. They also require the technical guidance to supply metrics or benchmarks for template components and to draw on voluntary consensus standards and industry best practices. For implementers this means NIST will aim to specify both what to disclose and how to measure or validate those disclosures, which matters to auditors and procurement teams seeking objective comparators.
Stakeholder collaboration and public comment
NIST must collaborate with a broad set of external actors—companies, universities, nonprofits, international standards bodies, and federal agencies—and it must publish a draft in the Federal Register with a public-comment period of not less than 60 days. That creates a predictable, notice-and-comment-style engagement process that gives industry and civil-society actors a formal window to influence definitions, metrics, and sector-specific tailoring of the template.
Reporting, publication, and potential permanence
Within 12 months of starting the pilot, NIST must deliver a report to specific House and Senate committees evaluating the pilot and, if judged effective, proposing a plan to make it permanent, including administration. The statute also requires NIST to post the template and guidelines on its public website. The near-term reporting deadline imposes a tight schedule for pilot design, stakeholder engagement, and drafting that NIST must meet if the pilot is to move toward a longer-term program.
Definitions
The bill provides statutory definitions for key terms like "Director," "artificial intelligence" (by reference to an existing federal definition), "artificial intelligence model," "institution of higher education," "nonprofit organization," and "international standards organization." These definitions narrow the bill’s scope and clarify who and what are covered when NIST conducts outreach and composes the template.
This bill is one of many.
Codify tracks hundreds of bills on Technology across all five countries.
Explore Technology in Codify Search →Who Benefits and Who Bears the Cost
Every bill creates winners and losers. Here's who stands to gain and who bears the cost.
Who Benefits
- Federal agencies that procure or operate AI: a standardized documentation template reduces due-diligence costs and provides repeatable inputs for procurement clauses, security reviews, and risk assessments.
- Procurement officers and compliance/audit teams in the private sector: consistent documentation makes technical reviews and third-party audits faster and more comparable across vendors.
- Standards organizations and testing labs: NIST’s guidance creates a common reference that can be adopted or adapted into conformance testing, certification criteria, or sector-specific standards.
Who Bears the Cost
- Model developers and vendors (especially small providers): preparing comprehensive documentation and associated benchmarks will require staff time, engineering work, and possibly independent validation, raising costs for development and release.
- NIST and other participating federal agencies: running a robust pilot, processing public comments, and maintaining published resources imposes administrative and technical workloads that rely on appropriations.
- Procurement and legal teams of buyer organizations: integrating the template into contracts, training staff to assess documentation, and aligning internal policies will require operational changes and possibly new tooling.
Key Issues
The Core Tension
The central tension is between maximizing transparency and auditability (which pushes toward detailed, standardized disclosures and measurable benchmarks) and protecting proprietary information, privacy, and the resource capacity of smaller developers (which pushes toward limited, flexible, or redacted disclosures). The bill attempts to thread this needle by making guidance voluntary and modular, but that very voluntariness risks becoming effectively mandatory if buyers or regulators adopt the template as a baseline.
The bill establishes a technical, voluntary resource rather than a direct regulatory mandate, but voluntary NIST resources can quickly become de facto requirements when adopted in procurement specifications or by regulators—raising questions about the downstream legal force of guidance. The statute conditions most activities on the availability of appropriations, creating implementation uncertainty: without funding, NIST’s pilot may be delayed or under-resourced, and the quality and breadth of stakeholder engagement could suffer.
Implementation design choices will also create trade-offs. Publishing detailed metrics and benchmarks improves auditability but can privilege vendors who can invest in test suites and measurement infrastructure, disadvantaging small or academic developers.
Requiring detailed provenance, datasets, or model internals in documentation increases transparency for auditors and researchers but raises intellectual property and privacy concerns that documentation design must mitigate (for example, by enabling redaction or synthetic summaries). Finally, selecting which voluntary standards to incorporate will be contentious: different sectors prioritize different risk metrics, and international alignment will require careful mapping of definitions and norms.
Try it yourself.
Ask a question in plain English, or pick a topic below. Results in seconds.