What does Preamble of HR 6356 do?

This bill, formally titled 'To establish protections for individual rights with respect to computational algorithms, and for other purposes,' aims to establish protections for individual rights related to computational algorithms. The bill was introduced in the House of Representatives on December 2, 2025.

What does SHORT TITLE; TABLE OF CONTENTS. of HR 6356 do?

This bill is formally titled the Artificial Intelligence Civil Rights Act of 2025.

What does Pre-deployment evaluations and post-deployment impact of HR 6356 do?

This section concerns pre-deployment evaluations and post-deployment impact assessments.

What does Study on explanations regarding the use of covered of HR 6356 do?

This section requires a study on explanations regarding the use of covered algorithms.

What does DEFINITIONS. of HR 6356 do?

This section defines key terms used throughout the Act. “Collect” and “collection” regarding personal data mean buying, renting, gathering, obtaining, receiving, accessing, or otherwise acquiring such data. A “commercial act” means an act conducted for monetary or other valuable consideration. The “Commission” refers to the Federal Trade Commission. “Consequential action” refers to an act likely to have a material effect on areas such as employment, education, housing, healthcare, and elections. “Covered algorithm” includes computational processes derived from machine learning or other complex techniques used in consequential actions. “Covered language” refers to the 10 most spoken languages in the United States. “De-identified data” is information that cannot be linked to an individual and requires specific technical and contractual safeguards. “Deployer” means any person that uses a covered algorithm for a commercial act. “Developer” means any person that designs or modifies an algorithm intended to be used as a covered algorithm. “Protected characteristic” includes traits like race, religion, disability, and age.

What does DISCRIMINATION. of HR 6356 do?

This section prohibits developers and deployers from using covered algorithms in ways that cause disparate impact, discriminate, or make goods, services, or opportunities unavailable based on protected characteristics related to consequential actions. Exceptions exist for self-testing and auditing to prevent discrimination, expanding applicant pools to increase diversity or redress historic discrimination, and conducting good faith security or other non-commercial research. The section also excludes private clubs as defined in the Civil Rights Act of 1964.

What does PRE-DEPLOYMENT EVALUATIONS AND POST-DEPLOYMENT IMPACT of HR 6356 do?

This section outlines requirements for pre-deployment evaluations and post-deployment impact assessments of covered algorithms. Developers and deployers must conduct preliminary evaluations to assess the plausibility of harm resulting from the algorithm's use. If a harm is plausible, a full pre-deployment evaluation, including an independent auditor evaluation, is required. The full evaluation includes a detailed review of the algorithm's design, methodology, training, testing, and potential for harm or disparate impact. Deployers also must engage an independent auditor to conduct a pre-deployment evaluation, focusing on the algorithm's manner of operation and proportionality to its planned use. Post-deployment assessments are required to monitor the algorithm's performance, identify potential harm or disparate impact, and determine if modifications are needed. Developers and deployers must report evaluations and assessments to the Commission and publish summaries on their websites. The Commission is authorized to promulgate rules to carry out these requirements.

What does COVERED ALGORITHM STANDARDS. of HR 6356 do?

This section establishes standards for developers and deployers of covered algorithms. Developers and deployers must take reasonable measures to prevent and mitigate harm identified through evaluations or assessments described in section 102. They are required to ensure independent auditors have necessary information for evaluations and assessments. Consultation with stakeholders, including impacted communities, is mandated before deploying, licensing, or offering a covered algorithm. Certification is required, stating that algorithm use is unlikely to cause harm, disparate impact, or deceptive practices, and that benefits outweigh harms. Algorithms must function at a reasonable performance level and consistently with advertised performance. Data used in the algorithm's lifecycle must be relevant and appropriate to its purpose. Use of the algorithm must not likely violate the Act. It is unlawful to engage in false, deceptive, or misleading advertising regarding covered algorithms. Developers cannot knowingly offer or license algorithms for uses not evaluated in pre-deployment evaluations. Deployers cannot knowingly use algorithms for un-evaluated actions unless they assume developer responsibilities.

What does RELATIONSHIPS BETWEEN DEVELOPERS AND DEPLOYERS. of HR 6356 do?

This section outlines responsibilities for developers and deployers of covered algorithms. Developers must provide deployers with information to demonstrate compliance with the Act, including reports and data necessary for evaluations. Developers must either allow deployer assessments or arrange for independent audits of their policies and practices. Contracts between developers and deployers must clearly define data processing procedures, instructions for data handling, the purpose of data collection, data types, processing duration, and rights/obligations. Contracts cannot relieve either party of obligations under the Act and prohibit combining data from different sources. Both developers and deployers retain the right to raise concerns with enforcement agencies. Developers must retain copies of contracts with deployers for 10 years. If a developer provides services to a government entity, they must follow instructions from the government entity as well.

What does HUMAN ALTERNATIVES AND OTHER PROTECTIONS. of HR 6356 do?

This section establishes rights related to covered algorithms and consequential actions. The Commission is required to create regulations within two years of enactment regarding a right to human alternatives for consequential actions. These regulations must consider factors like clear and conspicuous notices, effectiveness, the magnitude of actions, and potential harm. Developers and deployers are prohibited from conditioning individual rights through deceptive practices or manipulating user interfaces to impair autonomy. The Commission must also create regulations within two years regarding a mechanism for individuals to appeal consequential actions resulting from covered algorithms. Regulations for appeals must ensure clarity, proportionality, accessibility, and consider training for human reviewers.

Back to bills

Artificial Intelligence Civil Rights Act of 2025 — HR. 6356

Last updated: 12/2/2025 · Introduced: 12/2/2025

Author: Yvette Clarke (D-NY)

TL;DR (AI)

The Artificial Intelligence Civil Rights Act of 2025 prohibits developers and deployers of covered algorithms from discriminating or causing disparate impact based on protected characteristics such as race, religion, disability, and age.
Developers and deployers must conduct pre-deployment evaluations and post-deployment impact assessments to identify potential harm or disparate impact from covered algorithms.
The Act establishes standards for developers and deployers, including requirements for independent audits, stakeholder consultation, and certification that algorithm use is unlikely to cause harm.
Individuals have the right to human alternatives for consequential actions, and developers and deployers are prohibited from conditioning individual rights through deceptive practices.
The Act provides for enforcement by the Federal Trade Commission, state attorneys general, and private individuals, with penalties for violations and protections for whistleblowers.

119th Congressin committeeOfficial page

Verified Votes

No verified votes yet.

Community Votes

No community votes yet. Be the first!

Voting as guest · Sign in for verified votes