AI resume screening has moved from a novelty to infrastructure: tools from HireVue, Eightfold, LinkedIn, and Workday now sit between millions of job applications and the recruiters who used to read them by hand. The vendors describe these systems as faster and fairer than human screening. Regulators, courts, and researchers have spent the past several years testing that claim, and the legal landscape itself has shifted significantly — in ways that matter to anyone deploying or being screened by these tools in 2026. Here is what the tools actually do, what their makers claim, and where the documented risk sits today.
At a Glance
| Tool | What It Claims to Do | Primary Mechanism |
|---|---|---|
| HireVue | Screens candidates early in hiring using job-specific predictive models | One machine-learning model per job type, trained to identify who is likely to succeed compared to current employees in that role |
| Eightfold | Scans an entire talent network in seconds and ranks candidates by fit | Deep learning matched against stated role requirements, including “skill learnability” |
| LinkedIn Hiring Assistant | Evaluates every applicant for a role and surfaces top candidates to recruiters | AI agent connected to a company’s ATS via Recruiter System Connect; also generates applicant targeting criteria and follow-up messages |
| Workday AI screening | Helps recruiters process large applicant volumes faster | Machine-learning recommendations inside recruiting analytics; is the subject of an active federal discrimination lawsuit over how its screening functions operate |
How These Tools Actually Work
Despite different marketing language, most AI screening tools do one of two things: rank candidates against a job description, or predict which candidates most resemble people already succeeding in the role. HireVue’s own description of its process, published on its blog, says each of its assessments is a separate machine-learning model built for a specific job type, with the single task of identifying which candidates in a pool are most likely to succeed compared with current job holders — the output is handed to a human recruiter as decision support, not a final decision. Eightfold describes a similar logic on its own site: deep-learning models scan a company’s entire talent network “within seconds” and rank candidates by how closely their profile matches a role’s stated requirements, including a factor it calls “skill learnability” — a candidate’s estimated potential to learn a new skill. LinkedIn’s Hiring Assistant, per LinkedIn’s own 2026 product update pages, connects to a company’s applicant tracking system and evaluates every applicant for a given role, while separately generating AI-written candidate-targeting criteria and follow-up messages for recruiters. Workday’s own recruiting product page describes its AI capability narrowly, as “machine learning recommendations” inside analytics dashboards, rather than as a dedicated resume-screening engine — though, as covered below, how its screening functions actually operate is now the subject of active litigation.
What the Vendors Claim
Every vendor examined here claims some version of “faster and less biased than a human.” Eightfold’s own candidate-screening page describes its ranking as “fast, unbiased, and intuitive.” HireVue goes further, describing an internal testing loop it says it runs for every model: auditing training data for bias, testing predictions for adverse impact, removing factors shown to produce biased results, and retraining until the disparity is addressed — the company gives the example of disabling a speech-speed feature after testing showed it scored men higher than women. Workday, in materials cited in ongoing litigation, has said its technology assesses only job qualifications and that it tests its products for bias through an internal Responsible AI program. These are vendor self-reports rather than independent, published audits, and none of the three companies’ own pages reviewed for this article cite third-party verification of their bias-testing results.
The Documented Legal and Bias Risks
The clearest primary-source rule in this space is New York City’s Local Law 144. Per the New York City Department of Consumer and Worker Protection’s own page, employers and employment agencies may not use an “automated employment decision tool” unless it has been subject to an independent bias audit within the past year, and a summary of that audit must be made publicly available; the city’s Department of Consumer and Worker Protection began enforcing the law on July 5, 2023.
Federal guidance has moved in the opposite direction. The EEOC’s technical assistance documents on assessing adverse impact from AI and algorithmic hiring tools were removed from eeoc.gov in January 2025 and, as of a March 2026 review by the National Law Review, had not been restored or replaced — only a broken-link shell page on AI and the ADA remained. That doesn’t change the underlying law: Title VII and the federal Uniform Guidelines on Employee Selection Procedures still apply, and the EEOC’s current Strategic Enforcement Plan still lists AI-related discrimination as a priority, but the explanatory guidance employers used to assess their own tools is currently unavailable. Into that gap, at least four states have moved with their own rules: California’s automated-decision-system regulations under its Fair Employment and Housing Act took effect October 1, 2025; Illinois’s HB 3773 takes effect January 1, 2026, with a private right of action; Texas’s TRAIGA takes effect the same day but requires proof of discriminatory intent rather than disparate impact alone; and Colorado’s AI Act (SB 24-205), covering high-risk employment decisions, is set to take effect June 30, 2026.
The most closely watched court case is Mobley v. Workday. According to a June 2026 ruling reported by HCAMag, a federal judge in San Francisco allowed the proposed class action to proceed, rejecting Workday’s motion to dismiss most claims. The plaintiff, who is Black, over 40, and has a disability, alleges he was rejected from more than 100 jobs at companies using Workday’s screening software, often within minutes of applying, and that the software flagged applicants using proxy indicators for disability such as gaps in employment history. The court held that Workday can be treated as an “employer” under federal anti-discrimination law because it performs screening functions that client companies would otherwise handle themselves — a significant finding for any vendor selling AI screening as a service, not just Workday. Workday disputes the allegations and points to its internal bias-testing program; the case remains an unresolved proposed class action, not a final verdict, and court filings cited in that reporting reference roughly 1.1 billion job applications processed through the challenged tools over the relevant period.
What to Do About It
With federal guidance in flux and state rules arriving on different timelines, the practical burden has shifted toward independent verification rather than vendor assurance. Open-source and commercial bias-auditing frameworks exist specifically for this gap: IBM’s AI Fairness 360 (now maintained by the Linux Foundation) is an open-source toolkit with more than 70 fairness metrics and ten bias-mitigation algorithms for testing machine-learning models before and after deployment. Commercial platforms such as Fiddler and Credo AI offer similar model-monitoring and AI-governance capabilities aimed at enterprise compliance teams, though this article has not independently verified their specific feature claims against primary sources the way it did for AI Fairness 360, so treat vendor-specific capability claims from either company with the same caution applied to the hiring-tool vendors above. The throughline for any employer using these tools in 2026: a vendor’s internal bias-testing claim is not the same as an independent audit, and in New York City, only an independent audit satisfies the law.
Bottom Line
AI resume and candidate screening tools genuinely do speed up high-volume hiring, and every major vendor now markets some form of internal bias testing. But the regulatory floor supporting those claims is uneven and moving: NYC’s Local Law 144 is real, specific, and enforced; federal EEOC guidance that once helped employers interpret their Title VII obligations has been pulled with no replacement; and a live federal lawsuit is testing, in real time, whether a vendor’s screening software can itself be held liable for discrimination. Anyone deploying or relying on these tools should treat 2026 as a year of active legal uncertainty, not settled practice.
FAQ
Is AI resume screening illegal?
No, it is not illegal outright, but its use is increasingly regulated. New York City requires an independent bias audit before using an “automated employment decision tool,” and several states now impose their own disclosure or disparate-impact rules.
Did the EEOC say AI hiring tools cause discrimination?
Its prior technical assistance documents addressed how to assess adverse impact from such tools, but those pages were removed from eeoc.gov in January 2025 and have not been replaced as of a March 2026 legal review. The underlying Title VII law and Uniform Guidelines on Employee Selection Procedures remain in force regardless.
What is Mobley v. Workday?
It’s a proposed federal class action alleging Workday’s AI screening software disadvantaged applicants based on race, age, and disability. A judge allowed most claims to proceed in a June 2026 ruling, but the case has not been decided on the merits and Workday denies wrongdoing.
Can a bias-auditing tool guarantee a hiring AI is fair?
No single tool guarantees fairness. Frameworks like IBM’s open-source AI Fairness 360 provide metrics and mitigation algorithms to test models, but they measure specific statistical definitions of fairness and require ongoing use — they are a mitigation step, not a one-time certification.
Do all states now require bias audits for hiring AI?
No. As of this writing, only New York City has an enforced independent-audit requirement. California, Illinois, Texas, and Colorado have each passed their own, differently structured AI employment laws, with effective dates ranging from October 2025 through June 2026, but none mirror NYC’s audit mandate exactly.
Sources
- NYC Department of Consumer and Worker Protection — Automated Employment Decision Tools (Local Law 144)
- National Law Review — Federal Government Quietly Removed Its AI Hiring Guidance; Four States Are Writing
- HCAMag — Workday AI bias case moves forward, putting HR hiring tools on notice
- HireVue — HireVue Assessments and Preventing Algorithmic Bias
- Eightfold AI — Candidate Screening
- LinkedIn — 2026 LinkedIn Hiring Release, Wave 1
- AI Fairness 360 (Linux Foundation / originally IBM)
