An Indigenous and Deaf woman employed in customer service at Intuit's TurboTax received consistently positive evaluations from supervisors and customers. When she was encouraged to pursue a management role in spring 2024, the company subjected candidates to an AI video interview system. These platforms are documented to systematically underrate speakers who are not white or are deaf, yet Intuit declined to grant the accommodation she requested. She was ultimately not selected, with feedback suggesting she should "practice active listening."

A 51-year-old Black man with depression and anxiety submitted more than 100 job applications through Workday's AI-driven hiring platform after losing his position in 2017. Many applications required him to complete Workday assessments or personality evaluations. Despite his strong qualifications, he encountered swift rejections—frequently within hours of submission to Workday systems. He attributes this pattern to algorithmic discrimination based on his age, race, and mental health condition.

These two individuals exemplify discrimination in the modern era, and with approximately 98.4% of Fortune 500 companies deploying AI in recruitment, the scope of potential algorithmic bias is staggering. Yet lawmakers are preparing to worsen this situation.

A two-pronged attack on civil rights

The Trump administration has initiated a sweeping assault on civil rights enforcement as AI capabilities expand. Within days of taking office, federal agencies began dismantling existing AI safeguards and policies, with the Equal Employment Opportunity Commission and Department of Labor erasing workplace AI discrimination guidance from their sites. This marks merely one component of a larger effort to undermine civil rights safeguards that have protected Americans for generations.

Simultaneously, Congress is weighing language in a budget measure that would strip states of authority to regulate AI, removing any meaningful check on algorithmic discrimination without establishing federal alternatives. This absence of regulation creates a troubling dynamic: firms will gravitate toward discriminatory AI because it costs less than human judgment and proves difficult to challenge through existing legal mechanisms.

These two moves work in tandem. By first dismantling federal oversight and then blocking state action, they form a coordinated strategy to deny ordinary people fair treatment in employment, credit, and healthcare coverage. The outcome benefits corporations precisely when safeguards are most needed.

Why AI discrimination demands new legal frameworks

The explanation lies in understanding how discrimination statutes function in practice. Federal civil rights legislation was drafted before opaque algorithmic systems existed.

Title VII of the Civil Rights Act bars workplace discrimination. Claims proceed through two pathways. First, "disparate treatment" involves proving bias based on race, gender, or another safeguarded category; courts examine evidence of "animus" or intent to discriminate, such as a male manager expressing reluctance to supervise women, or a hiring committee member labeling Black applicants as DEI selections.

The second pathway is "disparate impact," addressing policies or practices that disproportionately affect individuals from a particular racial, gender, or other demographic group. Unlike disparate treatment claims, which demand proof of deliberate prejudice, disparate impact examines unnecessary obstacles, acknowledging that discrimination frequently operates through ostensibly neutral mechanisms. Many AI systems function precisely this way: applying identical criteria but producing unequal results.

Bringing a disparate impact claim requires pinpointing the exact hiring mechanism—such as deployment of an AI screening tool—generating the disproportionate outcome. Research demonstrates that AI screening tools frequently produce such effects: one study showed AI systems favored resumes with white-sounding names 85% of the time.

Consider the Deaf Indigenous applicant in Colorado. Nothing suggests the employer or the AI video software developer harbored conscious intent to exclude Deaf or Indigenous individuals. Yet the technology—and the employer's choice to deploy it—prevented qualified candidates like her from receiving equitable consideration. The employer may demonstrate that the video interview system serves operational needs, but if a less damaging option exists and is not being used, discrimination has occurred.

Disparate impact accountability was identified as a target in the billionaire-backed Project 2025, and President Trump has acted accordingly. He recently issued an executive order called "Restoring Equality of Opportunity and Meritocracy" designed to eliminate disparate impact protections throughout the federal government. While individuals retain the ability to pursue private disparate impact claims under Title VII and comparable statutes, they will wage this fight without federal agencies providing support.

Even assuming the assault on disparate impact does not succeed, the disparate impact framework alone may prove insufficient. Those harmed by discrimination must establish that the AI caused the injury, yet specialists themselves struggle to trace connections between inputs and outputs in sophisticated systems. Furthermore, most AI developers classify their platforms as proprietary, restricting access to the information required to demonstrate discrimination. Courts frequently demand such evidence early in proceedings, before plaintiffs obtain the materials necessary to construct their case.

In essence: American civil rights statutes were constructed around human judgment, not inscrutable algorithms. Without modernization, the protections Americans have depended on for decades will erode as opaque, unaccountable AI systems proliferate.

The need for state leadership

This reality explains why crafting fresh legislation at the state level has grown vital. The federal measure under consideration would prohibit states from enacting any AI regulations and would render existing state statutes moot, including elementary safeguards like those in Colorado. Colorado's statute mandates that those deploying high-risk AI systems exercise reasonable diligence to shield individuals from known or reasonably anticipated risks stemming from algorithmic bias.

Colorado's model does not impose broad AI prohibitions. Rather, it ensures workers and consumers learn when algorithms are making consequential decisions about them and supplies the transparency and documentation required to challenge discriminatory outcomes. These restrained measures exemplify the kind of creative, pragmatic approaches states can pursue when the federal government abdicates responsibility.

Going forward

AI holds tremendous promise. Medical research shows AI can assist physicians in predicting survival rates across numerous cancer categories and forecasting how patients will respond to conventional therapies. Educational applications demonstrate that AI systems can customize learning to individual needs and deliver one-on-one instruction to large populations. However, absent adequate governance, AI systems risk embedding and amplifying discrimination.

Congress must immediately discard this AI provision. The 40 bipartisan state attorneys general who labeled it "irresponsible" recognize the stakes: the methodical dismantling of civil rights safeguards through a false dichotomy between sensible oversight and technological advancement.

Should Congress decline to establish AI accountability standards, it forfeits the authority to obstruct states from defending their populations. As society harnesses AI's advantages, resistance to discrimination and commitment to equal access must remain non-negotiable.

Progress and civil rights protections are not mutually exclusive—but only if both are actively pursued.

Source: Tech Policy Press