Employers bear legal liability for AI-assisted hiring decisions even when algorithms are purchased or licensed from third-party vendors, according to analysis published June 9 by Rimon Law partner Tara Humma in HR Dive. The warning addresses a widespread misconception among HR teams that software creators shoulder discrimination risk when automated screening, interview analysis, or productivity monitoring tools produce unlawful employment outcomes.
TL;DR: Employers using AI hiring and workforce tools remain liable for discriminatory outcomes under federal disparate impact doctrine and expanding state regulations, regardless of whether algorithms were built in-house or licensed from vendors.
Federal Guidance Rescinded But Disparate Impact Doctrine Remains
The U.S. Equal Employment Opportunity Commission rescinded AI-specific guidance in early 2025 following a Trump administration executive order directing agencies to deprioritize disparate impact enforcement, Humma noted. The Biden-era guidance had cautioned employers about discrimination risks in AI-driven employment decisions, specifically warning that companies could face liability for discriminatory outcomes from third-party tools “even without an intent to discriminate.”
The rescission does not eliminate legal risk, according to the analysis. Federal court precedent upholding disparate impact discrimination claims—which focus on results rather than intent—remains binding law. “Neither an executive order nor rescission of agency guidance can overturn years of judicial precedent,” Humma wrote.
The disparate impact doctrine allows plaintiffs to challenge employment practices that produce discriminatory outcomes across protected classes, even when employers lack discriminatory intent. Courts have applied the framework to hiring tests, physical requirements, and background check policies for decades.
State Laws Fill Federal Enforcement Gap
Illinois now prohibits AI use that results in employment discrimination based on protected characteristics and requires employers to disclose when AI tools are in operation, according to the analysis. New York City mandates bias audits for AI tools used in employment decisions and requires notification to candidates and employees about algorithmic screening.
Several additional states have passed or proposed similar laws that either codify disparate impact protections or establish specific requirements for AI deployment in hiring and workforce management, Humma reported. The state-level activity contradicts assumptions that reduced federal EEOC enforcement signals lower compliance risk.
“State agencies and individuals can still bring these claims against employers under both state and federal law,” the analysis stated. Massachusetts regulators have opened legal scrutiny of AI screening tools over concerns that some products function as prohibited lie-detector tests under state statute.
HR Teams Lack Visibility Into AI Decision Logic
Many organizations deploy AI in applicant screening, interview evaluation, scheduling, productivity monitoring, promotion decisions, and disciplinary processes without understanding the underlying data sources, training methodologies, or whether protected characteristics influence outcomes indirectly, according to Humma’s assessment.
“This lack of understanding creates risk, and courts are unlikely to give credit to a defense that blames the creator or vendor if a claim arises,” she wrote. Managers in some companies follow AI-suggested outcomes without review, while other teams maintain human oversight protocols. The degree of human involvement directly affects legal exposure.
Harvard Business School and Accenture research documented that applicant tracking systems block millions of qualified candidates before human review occurs, often through algorithmic filters that correlate with protected class membership. Employers bear liability for those filtering decisions regardless of ATS vendor ownership.
Bias audits represent “one of the strongest defenses available if a claim comes up,” even in jurisdictions that do not mandate them, Humma advised. HR departments need documentation showing what controls exist to prevent discriminatory outcomes and which teams conduct oversight of algorithmic recommendations.
Compliance Steps for Recruiting and HR Leaders
HR teams should inventory where AI tools operate across the talent lifecycle, from resume parsing and knockout question logic to interview scheduling bots and performance monitoring dashboards, according to the analysis. Some employers inadvertently deploy age discrimination filters through hidden date-of-birth calculations in ATS configurations.
Understanding how managers interact with AI recommendations matters for liability assessment. Tools that auto-advance or auto-reject candidates without human review carry higher risk than advisory systems where recruiters manually confirm decisions. The distinction matters in litigation when plaintiffs argue employers delegated protected-class decision-making to unaudited algorithms.
Employers should evaluate whether tools produce disparate outcomes by protected class, even where state law does not mandate formal bias audits. The analysis recommends documenting audit protocols, control mechanisms, and human oversight procedures that demonstrate active management of discrimination risk rather than passive acceptance of vendor assurances.
Vendor contracts often disclaim liability for discriminatory outcomes, leaving employers fully exposed when algorithmic tools screen out protected-class candidates at disproportionate rates. “Just like with any other tool used by businesses, the employer may own the liability for decisions made with AI assistance,” Humma wrote.
Reading Between the Lines
The compliance gap matters acutely for mid-market and enterprise recruiting teams using modern ATS platforms with embedded AI candidate-ranking features. Many talent acquisition leaders treat algorithmic screening as a vendor-managed black box rather than an employer-owned hiring process subject to the same Title VII obligations as human recruiter decisions.
The practical implication: every AI-powered resume screen, video interview analyzer, or productivity monitor in the hiring workflow requires the same disparate impact analysis an employer would conduct for a paper-and-pencil employment test. Federal enforcement priorities may shift with administrations, but state regulators and private plaintiffs operate independently. Contracts that shift liability to software vendors rarely survive judicial scrutiny when discriminatory patterns emerge in candidate rejection data.
For HR leaders evaluating AI sourcing tools or interview chatbots, the compliance question isn’t whether the vendor conducted bias testing—it’s whether the employer can document how the tool was validated against protected-class outcomes in the specific context of the company’s candidate pool, job requirements, and decision-making workflow. Vendor assurances don’t transfer legal accountability.
