Eliminating Bias in Algorithmic Systems Act of 2026
- Sponsor
- Rep. Lee, Summer L. [D-PA-12]
- Committees
- Oversight and Government Reform Committee (primary)
- Last action
- Jan 15, 2026
Bottom line
The bill requires federal agencies to establish dedicated civil rights offices with AI expertise and report on algorithmic bias to ensure equitable outcomes in government technology use.
What it actually does
This bill mandates that all federal agencies that use, fund, or oversee algorithms establish or designate an office of civil rights. These offices must employ experts focused on identifying and mitigating bias, discrimination, and other harms caused by algorithms, particularly those affecting individuals based on protected characteristics. The bill also requires these offices to submit biennial reports to Congress detailing algorithmic risks, mitigation efforts, stakeholder engagement, and legislative recommendations, and establishes an interagency working group led by the Department of Justice.
Proponents argue
Supporters argue that this legislation is crucial for embedding civil rights protections directly into the federal government's use and oversight of artificial intelligence and algorithmic systems. They contend it will proactively address and mitigate the potential for these technologies to perpetuate or exacerbate existing societal biases and discrimination, ensuring that government services and opportunities are applied fairly to all individuals, especially those with protected characteristics.
Opponents contend
Critics might raise concerns that the bill could create significant bureaucratic burdens and costs for federal agencies, potentially slowing down the adoption of beneficial AI technologies. They may also argue that the broad definitions of 'covered algorithm' and 'protected characteristic' could lead to overreach or create regulatory uncertainty, and that finding sufficient numbers of experts with both civil rights and technical expertise will be challenging.
The bill is short and relatively straightforward, making it accessible for a quick read to understand its core mandates.
Section 1
Short Title
This section establishes the official short title of the Act as the "Eliminating Bias in Algorithmic Systems Act of 2026." This title provides a concise way to refer to the legislation and clearly indicates its primary purpose of addressing bias in algorithmic systems.
Supporters argue
Supporters argue that a clear and descriptive short title helps communicate the bill's purpose to the public and facilitates legislative discourse.
Critics contend
No specific opposition is typically raised against short titles, as they are standard legislative practice.
Tradeoffs
None apparent for this provision.
Section 2
Definitions of Key Terms
This section defines critical terms used throughout the Act, including 'agency,' 'covered agency,' 'covered algorithm,' and 'protected characteristic.' A 'covered agency' is any agency that uses, funds, oversees, or participates in the development of a 'covered algorithm.' A 'covered algorithm' is a complex computational process (like AI or machine learning) that could materially affect agency programs, economic opportunities, or protected rights. 'Protected characteristic' lists various traits such as race, sex, disability, age, and other federally protected classifications.
Section 3(a)
Mandate for Agency Civil Rights Offices on Algorithmic Bias
This subsection requires the head of each 'covered agency' to ensure their agency has an office of civil rights. These offices must employ experts and technologists specifically focused on identifying and mitigating bias, discrimination, and other harms caused by algorithms, particularly those affecting individuals based on protected characteristics. This institutionalizes a dedicated focus on algorithmic fairness within federal operations.
Section 3(b)
Biennial Reporting on Algorithmic Harms
This subsection mandates that each civil rights office established under the Act submit a report every two years to relevant congressional committees. These reports must detail the state of covered algorithms within the agency's jurisdiction, risks related to bias and discrimination, steps taken to mitigate harms, engagement with stakeholders (including industry, civil rights groups, and academics), and recommendations for legislative or administrative action. This ensures ongoing transparency and accountability.
Section 3(c)
Establishment of Interagency Working Group on AI and Civil Rights
This subsection requires the Assistant Attorney General for the Civil Rights Division of the Department of Justice to establish an interagency working group on covered algorithms and civil rights within one year of the Act's enactment. Each civil rights office established under subsection (a) will be a member of this working group, fostering coordination and shared learning across the federal government.
Section 3(d)
Authorization of Appropriations
This subsection authorizes the appropriation of "such sums as may be necessary" to each covered agency to carry out the provisions of this Act. This provides the legislative authority for future funding, allowing agencies to request and receive the financial resources needed to implement the bill's mandates, such as staffing civil rights offices and conducting reports, without directly appropriating funds.
Defines "covered algorithm" as a computational process using machine learning, natural language processing, artificial intelligence techniques, or other computational processing techniques of similar or greater complexity, which has the potential for a material effect on agency programs, economic opportunities, or protected rights.
Section 2(3)
Why it matters:While it is standard legislative practice to define terms upfront, the broad implications of this specific definition for the entire scope of the bill might not be immediately obvious to a casual reader.
Case for: Supporters argue that a broad and forward-looking definition is necessary to capture the evolving nature of AI and ensure that all potentially harmful algorithmic systems are covered, preventing loopholes as technology advances.
Case against: Critics might contend that the definition is too broad and vague, potentially encompassing systems not intended for regulation or creating uncertainty for agencies and developers about what precisely constitutes a "covered algorithm" and triggers compliance.
Estimated impact: This definition determines the entire scope of the bill's application, impacting potentially hundreds of federal algorithms and the agencies that use or oversee them, thereby shaping the landscape of AI governance across the federal government.
Defines "protected characteristic" by listing 16 specific traits (e.g., race, sex, disability, age, biometric information, source of income) and includes a catch-all for "Any other classification protected by Federal law." This comprehensive list expands the traditional scope of anti-discrimination efforts in the context of algorithms.
Section 2(4)
Why it matters:While defining terms is standard, the comprehensive and potentially expansive nature of this list, particularly the inclusion of certain modern categories, is a significant policy choice that could be overlooked if not carefully examined.
Case for: Proponents argue that a comprehensive list ensures broad protection against various forms of algorithmic discrimination, reflecting an inclusive understanding of civil rights in the digital age and addressing new vectors of bias.
Case against: Critics might argue that some inclusions, like "biometric information" or "income level" (with specific caveats), could expand the scope of anti-discrimination law in new ways that require more extensive public and legislative debate, or that the catch-all provision is overly broad.
Estimated impact: Significantly broadens the scope of anti-discrimination efforts to cover a wide range of personal attributes in algorithmic contexts, potentially requiring agencies to develop new assessment methods for these characteristics.