Who is Responsible When AI Messes Up? The Case for Federal AI Liability Standards.
Written by Aashvi Kenia
Legal Fragmentation
As artificial intelligence becomes more involved in decisions that shape people’s lives, Congress should act - the United States needs a comprehensive federal AI liability framework. Today, AI regulation in the United States is heavily fragmented, as businesses and consumers face over 1,000 AI-related bills introduced across state legislatures in 2025 alone, with over 700 the year before (Baker Botts, 2026). States like California, Texas, Colorado, and Illinois have each enacted their own AI governance rules, all with different timelines, requirements, and definitions. Meanwhile, on December 11, 2025, the Trump administration issued an Executive Order seeking to preempt state AI laws and to establish a "minimally burdensome national policy framework" to prioritize industry innovation over consumer protection (White House EO 14365, 2025). While the administration's instinct to create a unified standard is correct, it should not be a weak standard.
The Harm is Real
For example, AI bias is not a theoretical concern but a documented pattern. A landmark 2019 study found that a widely used clinical algorithm in hospitals was racially biased, as black patients had to be significantly sicker than white patients to receive the same care recommendation (Obermeyer, 2019). The algorithm had been trained on historical healthcare spending data, which reflected decades of wealth inequality. The result was an AI system that automated discrimination at scale (ACLU, 2025). In hiring, similar patterns were found. Amazon's AI recruiting tool exhibited gender bias because it was trained on a decade of predominantly male applicant data, systematically downgrading resumes from women. The tool was ultimately removed, but after it filtered out candidates who should have received interviews. A 2025 study in the Journal of Law and Society found that AI hiring systems create "feedback loops.” Once a biased early-career decision is made, that person's data trains the next model, compounding discrimination across generations (Sheard, 2025). These issues are not bugs, but systemic failures caused by biased training data and unchanging model design. When companies face no legal accountability for discriminatory actions, there is little incentive to fix them.
Legal Precedence
Currently, existing law is insufficient. While Title VII of the Civil Rights Act, the ADA, and the Fair Housing Act technically apply to AI decisions, enforcement is inconsistent and proving intent is difficult. The FTC's Section 5 authority prohibits unfair or deceptive practices, but guidance is limited and penalties are rare (Drata, 2026). This jumble of state laws, while well-intentioned, creates compliance nightmares for businesses and leaves consumers in some states without any protections.
What is needed is a federal AI liability statute modeled on existing product liability law. Just as a pharmaceutical company can be held liable for a defective drug even without proven intent to harm, AI developers and deployers should be liable for discriminatory or harmful outputs from their systems. Specifically, Congress should require mandatory algorithmic impact assessments before deploying AI in high-stakes contexts (employment, credit, healthcare, housing), transparency disclosures so affected individuals know when AI influenced a decision about them, independent third-party audits of high-risk AI systems, and a private right of action allowing individuals harmed by biased AI to seek legal relief.
Utah's Artificial Intelligence Policy Act already points in the right direction. It expressly blocks companies from avoiding liability by blaming the AI itself (National Law Review, 2025). In addition, New York City's Local Law 144 requires employers using automated decision tools in hiring to conduct bias audits and disclose their use. These are good models, but state-by-state implementation is not enough. Due to profit and innovation-related incentives, policymakers are slow to act. We need federal minimum standards that every company must meet, regardless of where they operate.
Rebuttal
Critics of AI regulation argue that liability rules will prevent innovation and put U.S. companies at a disadvantage. However, the EU AI Act, which classifies AI systems by risk level and imposes strict requirements on high-risk applications, demonstrates that regulation and a thriving AI sector can coexist. American companies already comply with the EU Act in European markets, so there is no reason they cannot comply with comparable domestic standards. Moreover, the cost of inaction is paid by workers who lose jobs to biased algorithms, patients who receive inferior care, and borrowers who are denied credit they qualify for, which are real economic harms.
Solution
Congress should pass a federal AI Accountability and Liability Act by the end of 2026. The bill should establish a tiered risk framework, require pre-deployment bias testing for high-risk AI in employment, healthcare, and credit, mandate transparency notices to affected individuals, and create a private right of action for AI-related discrimination. The FTC and EEOC should be granted explicit enforcement authority and adequate resources to investigate algorithmic harm.
Every day that Congress waits means someone's job application, loan approval, or medical referral is decided by a system that cannot be questioned and is not held accountable, and it is time to fix it.
Sources
American Civil Liberties Union. (2025). Algorithms Are Making Decisions About Health Care, Which May Only Worsen Medical Racism. https://www.aclu.org/news/privacy-technology/algorithms-in-health-care-may-worsen-medical-racism
All About AI. (2025). AI Bias Report 2025: LLM Discrimination Is Worse Than You Think. https://www.allaboutai.com/resources/ai-statistics/ai-bias/
Baker Botts. (January 2026). U.S. Artificial Intelligence Law Update: Navigating the Evolving State and Federal Regulatory Landscape. https://www.bakerbotts.com/thought-leadership/publications/2026/january/us-ai-law-update
Bryan Cave Leighton Paisner. (n.d.). US state-by-state artificial intelligence legislation snapshot [Map]. https://www.bclplaw.com/en-US/events-insights-news/us-state-by-state-artificial-intelligence-legislation-snapshot.html
Drata. (2026). Artificial Intelligence Regulations: State and Federal AI Laws 2026. https://drata.com/blog/artificial-intelligence-regulations-state-and-federal-ai-laws-2026
National Law Review. (2025). 2026 Outlook: Artificial Intelligence. https://natlawreview.com/article/2026-outlook-artificial-intelligence
Obermeyer, Z., Powers, B., Vogeli, C., & Mullainathan, S. (2019). Dissecting racial bias in an algorithm used to manage the health of populations. Science, 366(6464), 447–453. https://doi.org/10.1126/science.aax2342
Sheard, J. (2025). Algorithm-facilitated discrimination: a socio-legal study of the use by employers of artificial intelligence hiring systems. Journal of Law and Society. https://doi.org/10.1111/jols.12535
White House. (December 11, 2025). Ensuring a National Policy Framework for Artificial Intelligence. Executive Order No. 14,365. https://www.whitehouse.gov/presidential-actions/2025/12/eliminating-state-law-obstruction-of-national-artificial-intelligence-policy/
