Full bill text rendering is coming soon.
In the meantime, read the full text on Congress.gov ↗.
Regulatory or Legal Changes Top 5
Mandatory Model Safety Plan for AI Developers
Requires AI developers to create, publish, and follow a safety plan for every AI model or system they build or significantly modify. Each plan must include a risk assessment, mitigation measures, and a named responsible officer, and must be filed with the Secretary of Commerce.
- Population Scope Low Directly regulated population is AI developers who create or substantially modify AI systems — a narrow commercial sector, well under 10% of the U.S. population.
- Budgetary Magnitude Low No dollar amount, appropriation, or funding stream is authorized in this provision.
- Legal / Regulatory Depth High Creates a new affirmative statutory duty — developers must produce, publish, follow, and file a safety plan for each AI system — imposing a binding compliance obligation that did not previously exist in federal law.
- Degree of Discretion Granted Medium The Secretary of Commerce (DOC) determines the form and manner of plan submission, granting meaningful but bounded implementing authority over procedural requirements rather than substantive content.
- Implementation & Enforcement Burden High Imposes a recurring per-product affirmative compliance obligation on every covered developer for each system created or substantially modified, creating ongoing administrative load across the regulated population.
- Temporal Commitment High Persists indefinitely until Congress acts to repeal it.
No signal yet
Core Policy Mechanism Top 5
Public AI Incident and Flaw Database
Directs the National Institute of Standards and Technology (NIST) to create a publicly accessible database of AI safety and security incidents and flaws within one year, built from voluntary and mandatory reports and including causal analysis.
- Population Scope Medium Directly affects frontier AI developers and critical infrastructure operators through mandatory reporting duties, while the public database reaches researchers, civil society, and academic institutions broadly — a substantial but not population-wide share of affected entities.
- Budgetary Magnitude Low Same as above.
- Legal / Regulatory Depth High Establishes a new mandatory reporting regime with binding deadlines, a new public database infrastructure, and FOIA exemptions — a substantive restructuring of federal AI incident governance that did not previously exist in statute.
- Degree of Discretion Granted Medium National Institute of Standards and Technology (NIST) must establish reporting mechanisms and the database on fixed statutory deadlines, leaving format, prioritization, and nonbinding guidance content to NIST's discretion within those bounds.
- Implementation & Enforcement Burden Medium Mandatory reporting within 30 days (or 72 hours for urgent threats) creates recurring case-by-case compliance obligations for covered entities, and NIST must stand up and maintain new reporting and database infrastructure.
- Temporal Commitment High Same as above.
No signal yet
Core Policy Mechanism Top 5
Agentic AI Risk Profile and Documentation Standards
Requires the National Institute of Standards and Technology (NIST) to develop a risk management guide and a standardized documentation template specifically for AI systems that act with greater independence, helping organizations assess and disclose how these AI agents are designed and controlled.
- Population Scope Medium Targets organizations across all sectors and sizes that develop or deploy AI agents — a broad class of private, public, and academic entities, though adoption of nonbinding guidance is not mandated.
- Budgetary Magnitude Low Same as above.
- Legal / Regulatory Depth Medium Directs National Institute of Standards and Technology (NIST) to produce specific guidance deliverables on a statutory deadline — a new procedural mandate on agency conduct — but both outputs are explicitly nonbinding, limiting substantive legal effect on the regulated population.
- Degree of Discretion Granted Medium NIST's Director must produce the Agentic AI Profile and documentation template within fixed timeframes but retains broad discretion over content, methodology, and the collaborative process for developing both instruments.
- Implementation & Enforcement Burden Low Both deliverables are nonbinding guidance instruments, imposing no compliance obligations on the regulated population and only internal agency development burdens on NIST.
- Temporal Commitment High Same as above.
No signal yet
Core Policy Mechanism Top 5
Permanent AI Safety Board within Commerce
Creates a permanent board inside the Department of Commerce to address AI safety risks, with members drawn from federal agencies and outside experts in AI, industry, and consumer protection. The board can enter international agreements on AI safety.
- Population Scope Low Directly affects a small defined set — federal agency representatives, appointed nongovernmental experts, and international partners engaged through the Board — well under 10% of the U.S. population.
- Budgetary Magnitude Low Same as above.
- Legal / Regulatory Depth Medium Creates a new permanent advisory body with defined membership, security clearance requirements, and international agreement authority — new institutional legal infrastructure — but the Board is advisory, not a binding regulatory authority over the regulated population.
- Degree of Discretion Granted Medium The Secretary of Commerce (DOC) and Secretary of State retain discretion over the scope and terms of international AI safety agreements, while Board bylaws, quorum, and decision-making procedures are largely self-determined within statutory parameters.
- Implementation & Enforcement Burden Low Burdens are internal to the government — standing up the Board, sponsoring security clearances, and adopting bylaws — with no compliance obligations imposed on the broader regulated population.
- Temporal Commitment High Persists indefinitely until Congress acts to repeal it, explicitly exempted from FACA's automatic termination requirements.
No signal yet
Carve-outs, Exemptions, Eligibility Top 5
Exemption from disclosure; reports and anonymity
The Director of the Institute shall populate the database developed and established under paragraph (2)(A) with incidents and artificial intelligence flaws based on public reports and information shared using the mechanism established pursuant to paragraphs (1) and (3), ensuring that any incident…
- Population Scope Low Affects only entities that submit reports to the database and the parties named in those submissions — a narrow population of reporters and reported parties.
- Budgetary Magnitude Low Same as above.
- Legal / Regulatory Depth Low This component passed zero gates in LLM1's evaluation and lacks sufficient standalone operative content to assess; treated as L across all metrics.
- Degree of Discretion Granted Low No independent discretionary authority is assigned by this component as described.
- Implementation & Enforcement Burden Low No independent enforcement mechanism or compliance obligation is identifiable in this component as described.
- Temporal Commitment Low No independent operative commitment is established by this component as described.
No signal yet
No possible riders have been surfaced for this bill.
Summary
This bill sets new federal rules for managing safety and security risks from artificial intelligence (AI) systems. It applies mainly to companies that build or deploy AI, as well as operators of critical infrastructure that use AI. Among other elements, the bill * creates a permanent Artificial Intelligence Safety Board within the Department of Commerce (DOC) to advise on AI risks; * requires AI developers to write, publish, and file a safety plan for each AI system or model they build or significantly change; * directs the National Institute of Standards and Technology (NIST) to build a public database of AI safety and security incidents, with both voluntary and mandatory reporting; * mandates that developers of the most powerful AI models and critical infrastructure operators report confirmed incidents to NIST within 30 days, or 72 hours for urgent threats; * and directs NIST to develop risk management guidance and documentation standards for AI systems that act with greater independence, sometimes called AI agents.
AI-generated summary, pending human review.
| Version | Event | Date | User support | Your vote | Roll calls |
|---|---|---|---|---|---|
| Original |
Initial publication
Sep 29, 2026
|
Sep 29, 2026 | No votes yet | — | — |
Log in or create an account to record your vote on this legislation. Results are visible to all visitors.
Regulatory or Legal Changes Top 5
Mandatory Model Safety Plan for AI Developers
Requires AI developers to create, publish, and follow a safety plan for every AI model or system they build or significantly modify. Each plan must include a risk assessment, mitigation measures, and a named responsible officer, and must be filed with the Secretary of Commerce.
- Population Scope Low Directly regulated population is AI developers who create or substantially modify AI systems — a narrow commercial sector, well under 10% of the U.S. population.
- Budgetary Magnitude Low No dollar amount, appropriation, or funding stream is authorized in this provision.
- Legal / Regulatory Depth High Creates a new affirmative statutory duty — developers must produce, publish, follow, and file a safety plan for each AI system — imposing a binding compliance obligation that did not previously exist in federal law.
- Degree of Discretion Granted Medium The Secretary of Commerce (DOC) determines the form and manner of plan submission, granting meaningful but bounded implementing authority over procedural requirements rather than substantive content.
- Implementation & Enforcement Burden High Imposes a recurring per-product affirmative compliance obligation on every covered developer for each system created or substantially modified, creating ongoing administrative load across the regulated population.
- Temporal Commitment High Persists indefinitely until Congress acts to repeal it.
No signal yet
Regulatory or Legal Changes
Mandatory Model Safety Plan for AI Developers
Core Policy Mechanism Top 5
Public AI Incident and Flaw Database
Directs the National Institute of Standards and Technology (NIST) to create a publicly accessible database of AI safety and security incidents and flaws within one year, built from voluntary and mandatory reports and including causal analysis.
- Population Scope Medium Directly affects frontier AI developers and critical infrastructure operators through mandatory reporting duties, while the public database reaches researchers, civil society, and academic institutions broadly — a substantial but not population-wide share of affected entities.
- Budgetary Magnitude Low Same as above.
- Legal / Regulatory Depth High Establishes a new mandatory reporting regime with binding deadlines, a new public database infrastructure, and FOIA exemptions — a substantive restructuring of federal AI incident governance that did not previously exist in statute.
- Degree of Discretion Granted Medium National Institute of Standards and Technology (NIST) must establish reporting mechanisms and the database on fixed statutory deadlines, leaving format, prioritization, and nonbinding guidance content to NIST's discretion within those bounds.
- Implementation & Enforcement Burden Medium Mandatory reporting within 30 days (or 72 hours for urgent threats) creates recurring case-by-case compliance obligations for covered entities, and NIST must stand up and maintain new reporting and database infrastructure.
- Temporal Commitment High Same as above.
No signal yet
Core Policy Mechanism
Public AI Incident and Flaw Database
Core Policy Mechanism Top 5
Agentic AI Risk Profile and Documentation Standards
Requires the National Institute of Standards and Technology (NIST) to develop a risk management guide and a standardized documentation template specifically for AI systems that act with greater independence, helping organizations assess and disclose how these AI agents are designed and controlled.
- Population Scope Medium Targets organizations across all sectors and sizes that develop or deploy AI agents — a broad class of private, public, and academic entities, though adoption of nonbinding guidance is not mandated.
- Budgetary Magnitude Low Same as above.
- Legal / Regulatory Depth Medium Directs National Institute of Standards and Technology (NIST) to produce specific guidance deliverables on a statutory deadline — a new procedural mandate on agency conduct — but both outputs are explicitly nonbinding, limiting substantive legal effect on the regulated population.
- Degree of Discretion Granted Medium NIST's Director must produce the Agentic AI Profile and documentation template within fixed timeframes but retains broad discretion over content, methodology, and the collaborative process for developing both instruments.
- Implementation & Enforcement Burden Low Both deliverables are nonbinding guidance instruments, imposing no compliance obligations on the regulated population and only internal agency development burdens on NIST.
- Temporal Commitment High Same as above.
No signal yet
Core Policy Mechanism
Agentic AI Risk Profile and Documentation Standards
Core Policy Mechanism Top 5
Permanent AI Safety Board within Commerce
Creates a permanent board inside the Department of Commerce to address AI safety risks, with members drawn from federal agencies and outside experts in AI, industry, and consumer protection. The board can enter international agreements on AI safety.
- Population Scope Low Directly affects a small defined set — federal agency representatives, appointed nongovernmental experts, and international partners engaged through the Board — well under 10% of the U.S. population.
- Budgetary Magnitude Low Same as above.
- Legal / Regulatory Depth Medium Creates a new permanent advisory body with defined membership, security clearance requirements, and international agreement authority — new institutional legal infrastructure — but the Board is advisory, not a binding regulatory authority over the regulated population.
- Degree of Discretion Granted Medium The Secretary of Commerce (DOC) and Secretary of State retain discretion over the scope and terms of international AI safety agreements, while Board bylaws, quorum, and decision-making procedures are largely self-determined within statutory parameters.
- Implementation & Enforcement Burden Low Burdens are internal to the government — standing up the Board, sponsoring security clearances, and adopting bylaws — with no compliance obligations imposed on the broader regulated population.
- Temporal Commitment High Persists indefinitely until Congress acts to repeal it, explicitly exempted from FACA's automatic termination requirements.
No signal yet
Core Policy Mechanism
Permanent AI Safety Board within Commerce
Carve-outs, Exemptions, Eligibility Top 5
Exemption from disclosure; reports and anonymity
The Director of the Institute shall populate the database developed and established under paragraph (2)(A) with incidents and artificial intelligence flaws based on public reports and information shared using the mechanism established pursuant to paragraphs (1) and (3), ensuring that any incident…
- Population Scope Low Affects only entities that submit reports to the database and the parties named in those submissions — a narrow population of reporters and reported parties.
- Budgetary Magnitude Low Same as above.
- Legal / Regulatory Depth Low This component passed zero gates in LLM1's evaluation and lacks sufficient standalone operative content to assess; treated as L across all metrics.
- Degree of Discretion Granted Low No independent discretionary authority is assigned by this component as described.
- Implementation & Enforcement Burden Low No independent enforcement mechanism or compliance obligation is identifiable in this component as described.
- Temporal Commitment Low No independent operative commitment is established by this component as described.
No signal yet
Carve-outs, Exemptions, Eligibility
Exemption from disclosure; reports and anonymity
Incomplete vote
You have not made a selection on all required items:
Choose Neutral if you prefer not to commit. Leaving without completing will remove all your votes for this bill from the totals.
Maximum recommendations reached
You have already recommended the maximum number of components allowed on this bill. To recommend [new component], choose one of your current recommendations to unselect: