Can AI make hiring decisions

Can employers use AI to screen applicants without violating federal hiring laws?

Yes, but AI screening must still satisfy the same federal discrimination, validation, and accommodation rules that apply to any selection practice. The legal question is what the tool does to applicants, not whether the selector is human or automated.

Federal law does not ask whether the selector is human or machine. It asks what the selection practice does. Title VII's disparate-impact provision still turns on whether a plaintiff can identify a particular employment practice and whether the employer can show it is “job related for the position in question and consistent with business necessity”. The Uniform Guidelines still supply the familiar benchmark: a selection rate under four-fifths of the most-selected group is generally regarded by the Federal enforcement agencies as evidence of adverse impact. That framework was written for tests, not LLMs, but nothing in the statutory text creates an AI exception.

The ADA adds a separate layer. The EEOC's AI-and-ADA guidance distills the present federal problem into three recurring categories: tools that screen out disabled applicants, tools that fail to accommodate them, and tools that elicit or infer disability-related information the statute does not allow. That matters less for drafting job descriptions or summarizing interviews and more for systems that score facial expression, voice, eye contact, typing rhythm, or other traits that can become downstream proxies for disability.

The consensus from the employment bar is tighter than the headlines suggest. Fisher Phillips, Jackson Lewis, Littler, Ogletree, and Morgan Lewis all treat AI hiring as old employment law plus a new state/local patchwork, not as a new permission structure. Fisher Phillips describes a “patchwork of various state and local laws”. Jackson Lewis says employer liability for AI-assisted employment decisions remains “anchored in long-standing civil rights laws”. On the core federal point, there is very little daylight between them.

Sources for this answer

Primary law

A.1 42 U.S.C. § 2000e-2(k)

Supports the cited proposition. (42 U.S.C. § 2000e-2(k))

job related for the position in question and consistent with business necessity

See 42 U.S.C. § 2000e-2(k).

Primary law

A.2 29 C.F.R. § 1607.4

Employers are required to maintain records disclosing the impact of selection procedures on protected groups, and federal agencies may infer adverse impact if such data is not maintained or if selection rates fall below the four-fifths rule.

Each user should maintain and have available for inspection records or other information which will disclose the impact which its tests and other selection procedures have upon employment opportunities of persons by identifiable race, sex, or ethnic group

See 29 C.F.R. § 1607.4.

Primary law

A.3 42 U.S.C. § 12112 and EEOC, Artificial Intelligence and the ADA

The EEOC provides guidance on how the Americans with Disabilities Act applies to the use of software, algorithms, and artificial intelligence in employment decision-making processes.

The Americans with Disabilities Act and the Use of Software, Algorithms, and Artificial Intelligence to Assess Job Applicants and Employees

See 42 U.S.C. § 12112 and EEOC, Artificial Intelligence and the ADA.

Law-firm commentary

A.4 Jackson Lewis commentary

The EEOC and DOJ have issued guidance regarding the use of artificial intelligence and algorithmic decision-making tools in the workplace, emphasizing that while these documents do not create new legal obligations, they clarify how existing federal civil rights laws, including the ADA, apply to these technologies.

The EEOC’s TAD applies the Americans with Disabilities Act (ADA), including regulations and existing guidance, where technology intersects with workplace legal issues.

See Jackson Lewis, EEOC, DOJ Release Expectations on Employers’ Use of Technology, AI for Employment Decisions.

Law-firm commentary

A.5 Fisher Phillips commentary

Supports the cited proposition. (Fisher Phillips commentary)

patchwork of various state and local laws

See Fisher Phillips, Comprehensive Review of AI Workplace Law and Litigation as We Enter 2025.

Law-firm commentary

A.6 Jackson Lewis commentary

Supports the cited proposition. (Jackson Lewis commentary)

anchored in long-standing civil rights laws

See Jackson Lewis, Trump’s AI EO: Reducing Regulatory Fragmentation Not Employer Responsibility.

Which state rules apply when AI hiring tools screen NYC, Illinois, or Colorado applicants?

NYC, Illinois, and Colorado can each attach different obligations to the same AI hiring workflow. The triggers turn on the tool, the job location or applicant location, and whether the system substantially affects employment decisions.

The first local rule that changed employer behavior at scale was New York City Local Law 144. Its trigger is not AI in the abstract. It is an AEDT that issues a “score, classification, or recommendation” and is used to substantially assist or replace discretionary decision making in hiring or promotion. Once that trigger is met, the statute requires a recent bias audit, public posting of a summary, and at least ten business days' notice before use. The DCWP materials also make two practical points. The law can reach some remote roles tied to a NYC office, and a completed bias audit is not a city-issued safe harbor from other discrimination law.

Illinois now works in two layers. The narrower one is the Artificial Intelligence Video Interview Act. It applies when AI analyzes recorded video interviews for “positions based in Illinois”, and it requires notice, an explanation of how the AI works and the general characteristics it uses, and applicant consent before the interview proceeds. The broader one is the Illinois Human Rights Act amendment effective January 1, 2026. That amendment moves Illinois beyond recorded-video analysis and into ordinary employment decision-making by prohibiting AI use that has a discriminatory effect, barring zip code as a proxy for a protected class, and defining AI broadly enough to include generative AI.

Colorado's AI Act is broader still, but timing matters. SB 24-205 reaches high-risk systems that make, or are a substantial factor in making a consequential decision about employment or an employment opportunity involving a Colorado resident. The duties in the enacted statute are systemic: reasonable care, risk management, impact assessment, notice, and explanation. But as of April 20, 2026, the central deployer obligations are not yet operative because SB25B-004 pushed the effective date to June 30, 2026. That makes Colorado important and slightly easy to overstate at the same time.

Illinois is where the law-firm commentary shows the clearest shift over time. Older summaries could still describe Illinois as mainly a video-interview jurisdiction. Newer commentary from Littler and Morgan Lewis no longer does. They treat the 2026 Human Rights Act amendment as the bigger development because it changes Illinois from a narrow interview law into a broader employment-AI rule about discriminatory effect and notice.

Ogletree's Colorado writing is useful for a different reason. It underscores that the act is designed to reach employment use, but also flags the under-50-employee exemption and the ambiguity in substantial factor. That combination is typical of the whole field: the direction of travel is clear, while the exact line between assistive and decision-making uses is still fuzzy.

Sources for this answer

Primary law

B.1 New York City Local Law 144 / Int. No. 1894-A

Supports the cited proposition. (New York City Local Law 144 / Int. No. 1894-A)

score, classification, or recommendation

See New York City Local Law 144 / Int. No. 1894-A.

Primary law

B.2 New York City Local Law 144 and DCWP AEDT materials

New York City Local Law 144 mandates that employers and employment agencies conduct bias audits of automated employment decision tools and provide specific notices to candidates before using such tools.

prohibits employers and employment agencies from using an automated employment decision tool unless the tool has been subject to a bias audit within one year of the use of the tool, information about the bias audit is publicly available, and certain notices have been provided to employees or job candidates.

See New York City Local Law 144 and DCWP AEDT materials.

Primary law

B.3 DCWP, Automated Employment Decision Tools (AEDT) and FAQPDF

New York City's Local Law 144 of 2021 mandates that employers and employment agencies conduct annual bias audits by an independent auditor and provide notice to candidates before using automated employment decision tools.

The Law prohibits employers and employment agencies from using an automated employment decision tool (AEDT) in New York City unless they ensure a bias audit was done and provide required notices.

See DCWP, Automated Employment Decision Tools (AEDT) and FAQ.

Primary law

B.11 820 ILCS 42/5

Supports the cited proposition. (820 ILCS 42/5)

positions based in Illinois

See 820 ILCS 42/5.

Primary law

B.4 820 ILCS 42, Artificial Intelligence Video Interview Act

The Artificial Intelligence Video Interview Act imposes specific disclosure, consent, data sharing, and deletion requirements on employers who use artificial intelligence to analyze video interviews of job applicants in Illinois.

An employer that asks applicants to record video interviews and uses an artificial intelligence analysis of the applicant-submitted videos shall do all of the following when considering applicants for positions based in Illinois before asking applicants to submit video interviews

See 820 ILCS 42, Artificial Intelligence Video Interview Act.

Primary law

B.5 Illinois Public Act 103-0804 and 775 ILCS 5/2-101

The Illinois Human Rights Act prohibits employers from using artificial intelligence in employment decisions that results in discrimination against protected classes or using zip codes as a proxy for such discrimination, and requires employers to notify employees when artificial intelligence is used for these purposes.

For an employer to use artificial intelligence that has the effect of subjecting employees to discrimination on the basis of protected classes under this Article or to use zip codes as a proxy for protected classes under this Article.

See Illinois Public Act 103-0804 and 775 ILCS 5/2-101.

Primary law

B.6 Colorado SB24-205

Colorado SB24-205 establishes duties for developers and deployers of high-risk artificial intelligence systems to use reasonable care to prevent algorithmic discrimination, with enforcement authority granted exclusively to the Attorney General.

A DEVELOPER OF A HIGH-RISK ARTIFICIAL INTELLIGENCE SYSTEM SHALL USE REASONABLE CARE TO PROTECT CONSUMERS FROM ANY KNOWN OR REASONABLY FORESEEABLE RISKS OF ALGORITHMIC DISCRIMINATION ARISING FROM THE INTENDED AND CONTRACTED USES OF THE HIGH-RISK ARTIFICIAL INTELLIGENCE SYSTEM.

See Colorado SB24-205.

Primary law

B.7 Colorado SB24-205 and SB25B-004

Colorado Senate Bill 25B-004 extends the effective date for the consumer protection requirements for algorithmic systems originally established by Senate Bill 24-205 to June 30, 2026.

In 2024, the general assembly enacted Senate Bill 24-205, which created consumer protections in interactions with artificial intelligence systems.

See Colorado SB24-205 and SB25B-004.

Law-firm commentary

B.8 Littler commentary

In the absence of comprehensive federal regulation, state and local jurisdictions are increasingly enacting legislation that imposes duties of reasonable care, transparency, and bias auditing on employers using artificial intelligence in employment decisions.

In the absence of federal regulation, several states have either passed or are considering legislation aimed at mitigating the risk of an employer’s use of an AI system resulting in algorithmic discrimination.

See Littler, What Does the 2025 Artificial Intelligence Legislative and Regulatory Landscape Look Like for Employers?.

Law-firm commentary

B.9 Morgan Lewis commentary

Employers integrating artificial intelligence into their workplace operations face significant legal risks and regulatory challenges, necessitating proactive compliance measures and transparent management practices.

the adoption of AI technology in the workplace also brings several new legal complexities and risks.

See Morgan Lewis, Artificial Intelligence in Employment: Key Takeaways.

Law-firm commentary

B.10 Ogletree Deakins commentary

The Colorado AI Act, effective February 1, 2026, imposes comprehensive compliance obligations on employers using high-risk AI systems in employment decisions, including requirements for risk management, annual impact assessments, and consumer notifications.

Colorado becomes the first U.S. state to enact comprehensive legislation regulating the use and development of AI systems.

See Ogletree Deakins, Colorado’s Artificial Intelligence Act: What Employers Need to Know.

How much human review is enough for AI hiring recommendations?

Human review helps only if it is real enough to change the outcome and explain the hiring decision. A rubber-stamp approval step will not by itself avoid federal, state, local, or vendor-policy risk.

For Claude specifically, the statutory and contractual stories point the same way. Anthropic's Usage Policy classifies employment decisions, resume screening, and hiring tools as high-risk uses and says “a qualified professional in that field must review the content or decision prior to dissemination or finalization”. Anthropic's own candidate guidance describes Claude as useful for job descriptions, interview questions, candidate communications, metrics, transcription, and sourcing, but says plainly: “We don't use your data to train Claude or let Claude make hiring decisions.”

The non-obvious consequence is that the legal boundary is not AI versus no AI. It is selection system versus support tool. A model used to draft job descriptions, propose interview questions, transcribe calls, or summarize recruiting metrics fits much more naturally inside the assistive pattern Anthropic itself describes. A model used to gate who advances, who gets rejected, or who receives an offer turns into a selection practice, and often into a regulated selection tool.

Human in the loop is not magic words. If the model still supplies the score, classification, or recommendation that substantially assists the decision under NYC law, or is a substantial factor under Colorado's act, the local statute can still apply. Federal law is even less impressed by labels. Title VII and the ADA look at effect, validation, screen-out, and accommodation, not at whether the product team inserted a human approval button at the end.

A bias audit is not the end of the story. New York City requires one, but the city's own materials say the law does not itself prescribe a particular remedial step after the audit. That means the audit is a threshold condition for use, not a declaration that the workflow is lawful overall. A tool can therefore be bias-audited and still difficult to defend under disparate-impact or ADA theories.

One view, reflected in some LL144 commentary, is that a genuinely independent human decision-maker may keep a workflow outside certain local triggers or at least weaken causation arguments. The counterview is that a rubber-stamp reviewer does not cure disparate impact, validation gaps, or ADA screen-out. No bright-line rule in the source set resolves that boundary.

Morgan Lewis and Jackson Lewis converge on one practical point. Human review matters, but not as a slogan. The review has to be real enough that the company can still explain job-relatedness, validation, accommodation, and why the human decision-maker was not just accepting a machine-generated ranking as fate. None of the firm commentary in the source set treats a vendor bias audit or model card as the end of the inquiry.

The more interesting disagreement is about scope at the margins. Littler emphasizes that NYC Local Law 144 only reaches tools that substantially assist or replace discretion, and Fisher Phillips goes further, suggesting the law may be narrower than its headlines if managers retain the predominant decision-making role. That is not disagreement about whether AI hiring is regulated. It is disagreement about how much real human judgment is enough to keep a particular workflow outside one city statute.

Sources for this answer

Vendor documentation

C.1 Anthropic, Usage Policy

Supports the cited proposition. (Anthropic, Usage Policy)

a qualified professional in that field must review the content or decision prior to dissemination or finalization

See Anthropic, Usage Policy.

Vendor documentation

C.2 Anthropic, Guidance on Candidates' AI Usage / How Anthropic uses Claude for hiring

Supports the cited proposition. (Anthropic, Guidance on Candidates' AI Usage / How Anthropic uses Claude for hiring)

We don't use your data to train Claude or let Claude make hiring decisions.

See Anthropic, Guidance on Candidates' AI Usage / How Anthropic uses Claude for hiring.

Primary law

C.3 New York City Local Law 144 / Int. No. 1894-A

Supports the cited proposition. (New York City Local Law 144 / Int. No. 1894-A)

score, classification, or recommendation

See New York City Local Law 144 / Int. No. 1894-A.

Primary law

C.4 Colorado SB24-205

A DEVELOPER OF A HIGH-RISK ARTIFICIAL INTELLIGENCE SYSTEM SHALL USE REASONABLE CARE TO PROTECT CONSUMERS FROM ANY KNOWN OR REASONABLY FORESEEABLE RISKS OF ALGORITHMIC DISCRIMINATION ARISING FROM THE INTENDED AND CONTRACTED USES OF THE HIGH-RISK ARTIFICIAL INTELLIGENCE SYSTEM.

See Colorado SB24-205.

Primary law

C.5 42 U.S.C. § 2000e-2(k)

Supports the cited proposition. (42 U.S.C. § 2000e-2(k))

job related for the position in question and consistent with business necessity

See 42 U.S.C. § 2000e-2(k).

Primary law

C.6 42 U.S.C. § 12112 and EEOC, Artificial Intelligence and the ADA

The EEOC provides guidance on how the Americans with Disabilities Act applies to the use of software, algorithms, and artificial intelligence in employment decision-making processes.

The Americans with Disabilities Act and the Use of Software, Algorithms, and Artificial Intelligence to Assess Job Applicants and Employees

See 42 U.S.C. § 12112 and EEOC, Artificial Intelligence and the ADA.

Primary law

C.7 DCWP, Automated Employment Decision Tools (AEDT) and FAQPDF

The Law prohibits employers and employment agencies from using an automated employment decision tool (AEDT) in New York City unless they ensure a bias audit was done and provide required notices.

See DCWP, Automated Employment Decision Tools (AEDT) and FAQ.

Primary law

C.8 29 C.F.R. § 1607.4

Each user should maintain and have available for inspection records or other information which will disclose the impact which its tests and other selection procedures have upon employment opportunities of persons by identifiable race, sex, or ethnic group

See 29 C.F.R. § 1607.4.

Law-firm commentary

C.9 Fisher Phillips commentary

Supports the cited proposition. (Fisher Phillips commentary)

patchwork of various state and local laws

See Fisher Phillips, Comprehensive Review of AI Workplace Law and Litigation as We Enter 2025.

Law-firm commentary

C.10 Jackson Lewis commentary

The EEOC’s TAD applies the Americans with Disabilities Act (ADA), including regulations and existing guidance, where technology intersects with workplace legal issues.

See Jackson Lewis, EEOC, DOJ Release Expectations on Employers’ Use of Technology, AI for Employment Decisions.

Law-firm commentary

C.11 Morgan Lewis commentary

the adoption of AI technology in the workplace also brings several new legal complexities and risks.

See Morgan Lewis, Artificial Intelligence in Employment: Key Takeaways.

Law-firm commentary

C.12 Littler commentary

In the absence of federal regulation, several states have either passed or are considering legislation aimed at mitigating the risk of an employer’s use of an AI system resulting in algorithmic discrimination.

See Littler, What Does the 2025 Artificial Intelligence Legislative and Regulatory Landscape Look Like for Employers?.

Can an AI hiring vendor be liable for discriminatory screening decisions?

Possibly, especially when plaintiffs argue that the employer delegated screening decisions to the vendor software. The source set does not identify an appellate merits ruling that settles how far that theory can travel.

The Workday litigation is testing whether an AI hiring vendor can be treated as an employer's agent when the employer delegates screening to the software. Plaintiffs say the vendor is part of the employment decision. Vendors say they provide configurable software while the employer owns the criteria and outcome. The source set does not identify an appellate merits holding settling that issue.

Sources for this answer

Law-firm commentary

D.1 Fisher Phillips commentary

The Mobley v. Workday decision highlights that employers and AI vendors face potential liability for disparate impact discrimination in algorithmic hiring tools, even in the absence of intentional bias.

the ruling serves as a warning to employers and AI vendors alike that they can be held accountable for algorithmic screening tools if they disproportionately harm protected groups – even if the bias wasn’t intentional.

See Fisher Phillips, Discrimination Lawsuit Over Workday’s AI Hiring Tools Can Proceed as Class Action: 6 Things Employers Should Do After Latest Court Decision.

Does an AI applicant score trigger FCRA consumer-report duties?

It can become an FCRA issue when a vendor compiles, enriches, or scores applicant data in a way plaintiffs characterize as consumer reporting. The Eightfold litigation shows the theory, but the source set does not resolve how broadly it will apply.

The Eightfold litigation is testing whether AI-generated match scores can trigger FCRA-style duties when a vendor compiles or enriches applicant data. Perhaps that theory stays confined to the public-data-enrichment end of the market. Perhaps it spreads further. Either way, it shows that automated hiring can pick up procedural obligations from outside employment discrimination law itself.

Sources for this answer

Law-firm commentary

E.1 Jones Walker commentary

The recent litigation against AI hiring vendors, such as the Eightfold case, demonstrates a shift toward framing automated screening tools as consumer reporting agencies, thereby allowing plaintiffs to bypass the difficult burden of proving algorithmic bias by focusing on procedural FCRA violations.

The Eightfold case isn't another AI discrimination lawsuit. It's a consumer protection action that reframes how plaintiffs can attack automated hiring.

See Jones Walker, AI Hiring Under Fire: What the Eightfold Lawsuit Means for Every Employer Using Algorithmic Screening.

Which location rules apply when AI hiring tools screen remote applicants?

Remote hiring can trigger multiple AI hiring rules because NYC, Illinois, and Colorado use different geographic hooks. A centralized funnel may need to account for office ties, role location, and applicant residence at the same time.

The same workflow can also attract different rules for different reasons. A remote job tied to a NYC office, a recorded video interview for an Illinois-based role, and a Colorado resident in the applicant pool can all cause different laws to attach to the same funnel. The practical result is not one national AI hiring rule. It is one hiring stack that keeps changing legal character as geography and workflow stage change.

The smallest companies do not necessarily escape just because the federal statutes are not written for every employer. Colorado has a narrow under-50 exemption on stated conditions. But the source set does not surface an express small-employer carve-out in NYC Local Law 144 or in Illinois's video-interview statute. So too small to matter is not a stable conclusion once state and local rules enter the picture.

DCWP says some fully remote jobs tied to a NYC office are covered. Illinois keys the video-interview law to positions based in Illinois. Colorado keys its act to consequential decisions about Colorado residents. The hard cases are centralized recruiting teams running one workflow across all three. The statutes overlap, but they do not share one geographic key.

Sources for this answer

Primary law

F.1 DCWP, Automated Employment Decision Tools (AEDT) and FAQPDF

The Law prohibits employers and employment agencies from using an automated employment decision tool (AEDT) in New York City unless they ensure a bias audit was done and provide required notices.

See DCWP, Automated Employment Decision Tools (AEDT) and FAQ.

Primary law

F.2 820 ILCS 42, Artificial Intelligence Video Interview Act

An employer that asks applicants to record video interviews and uses an artificial intelligence analysis of the applicant-submitted videos shall do all of the following when considering applicants for positions based in Illinois before asking applicants to submit video interviews

See 820 ILCS 42, Artificial Intelligence Video Interview Act.

Primary law

F.3 Colorado SB24-205

A DEVELOPER OF A HIGH-RISK ARTIFICIAL INTELLIGENCE SYSTEM SHALL USE REASONABLE CARE TO PROTECT CONSUMERS FROM ANY KNOWN OR REASONABLY FORESEEABLE RISKS OF ALGORITHMIC DISCRIMINATION ARISING FROM THE INTENDED AND CONTRACTED USES OF THE HIGH-RISK ARTIFICIAL INTELLIGENCE SYSTEM.

See Colorado SB24-205.

Primary law

F.4 New York City Local Law 144 / Int. No. 1894-A

Supports the cited proposition. (New York City Local Law 144 / Int. No. 1894-A)

score, classification, or recommendation

See New York City Local Law 144 / Int. No. 1894-A.