This article explores the growing role of artificial intelligence in hiring and workforce decisions—and the legal, compliance, and bias risks employers should understand. Learn how AI can create unintended adverse outcomes, why vendor technology doesn’t remove employer responsibility, what meaningful human oversight looks like, and practical steps organizations can take to use AI-enabled employment tools more responsibly.
Glossary of Key Terms
Adverse Impact: A result that appears neutral but affects one protected group more negatively than another.
Algorithmic Bias: A pattern in which an automated system produces unfair or unequal results. Bias may come from historical data, the system’s design, the information selected for analysis, or the way an employer uses the result.
Bias Audit: An evaluation of an automated tool’s results to determine whether the tool creates unequal outcomes for certain groups. Some jurisdictions require specific types of bias audits.
Machine Learning: A type of AI that identifies patterns in data and uses those patterns to make predictions or recommendations.
Proxy Variable: Information that may indirectly reveal or closely relate to a protected characteristic. For example, a ZIP code could sometimes act as a proxy for race or economic status.
Artificial intelligence (AI) is becoming an integral part of the employment process, as employers have used AI in increasing measure to find candidates, screen resumes, rank applicants, evaluate interviews, measure performance, and support workforce planning.
The tricky part: In some cases, those same employers may not realize it. Features that look inherent to their applicant tracking system, recruiting platform, scheduling tool, or performance-management program may, in fact, be the handiwork of AI.
These tools can save time and help teams manage large amounts of information. However, the more influence AI has over an employment decision, the more risk it may create.
AI Is Becoming Part of the Hiring Process
Artificial intelligence in hiring can appear at almost every stage of the employee lifecycle, including the potential candidate search, interview scheduling, pre-employment background check, and even layoff support. But not all uses carry the same level of risk.
For instance, a tool that sends an interview reminder is different from one that gives an applicant a score, and a system that organizes documents is different from one that automatically removes a candidate from consideration.
Employers should always look beyond whether a product is described as “AI-powered.” The more important question is what the tool does and how much it influences the final decision.
Recent AI Lawsuit Shows How AI Inputs Can Create Employment Risk
A recent lawsuit involving a major technology company is a timely example of the risks employers may face when they carelessly use AI for workforce decisions.
In July 2026, 26 employees filed a lawsuit alleging that AI-assisted performance measures, activity monitoring, AI token-use dashboards, and algorithmic rankings were used in connection with layoff decisions.
The plaintiffs claimed that the process disadvantaged employees who took medical, parental, family, pregnancy, caregiving, or bereavement leave, alleging that people could not produce the same activity or performance numbers while they were away from work or working with an accommodation.
The major technology company disputed the claims, arguing that the workforce and organizational decisions were made by people rather than AI. (The allegations have not been proven and should be treated as claims made by the plaintiffs.)
Why the Case Matters for Employers
The case highlights a risk that can arise even when a metric appears neutral.
Imagine a system that tracks completed tasks, time online, sales calls, keystrokes, or another form of activity. On the surface, it may measure everyone in the same way. However, equal measurement does not always mean equal treatment.
To use hypotheticals from the case: An employee on protected leave may have fewer recorded activities, or an applicant with a disability may interact differently with an online assessment, or a worker using an approved accommodation may complete tasks in a different way. Without context, an automated system may treat these differences as signs of poor performance or low potential.
This example should not be taken to mean that every performance metric or AI system is unlawful. Rather, it's a cautionary tale: employers must consider whether the information given by AI tools is complete, job-related, and properly understood before using it to make an important employment decision.
Why AI Hiring Decisions Create Risk
AI Can Repeat Bias Found in Existing Data
AI systems often learn from past information, which may include previous hiring decisions, employee performance ratings, promotion records, or other workforce data.
However, even historical data is not automatically fair or job-related: The data, while apparently neutral, may reflect years of unequal access, inconsistent management practices, or subjective decisions.
Potential examples of data that can inadvertently encourage discriminatory practices when used in a vacuum include the following:
Employment gaps caused by medical or family leave
Lower activity measurements during parental leave
Disability-related differences in test performance
ZIP codes in historical underserved areas
School names
Communication styles
Schedule availability
Access to certain technology
Nontraditional career paths
When a system learns from those patterns, it may repeat them at a larger scale. And when used without question, an employer could believe the system is removing human bias when it is actually turning old patterns into a repeatable process.
Employers May Not Understand the Result
Some AI hiring tools provide a clear explanation of what they measure. Others offer only a score, recommendation, or risk level. The opacity of each AI tool will vary wildly, depending on the type of model, the vendor, the available documentation, and the specific configuration of a tool.
This discrepancy is sometimes called the "black-box problem."
A decision becomes harder to defend, review, or correct when no one can explain how that decision was reached. Warning signs of a potential black-box problem may include the following:
The employer cannot explain why a candidate received a score
The vendor will not explain which types of data affect the result
The variables or weight given to them are unclear
The tool changes without proper notice or documentation
Hiring managers cannot review the underlying information
Candidates have no way to correct an error
Human Review Can Become a Rubber Stamp
Many vendors and employers point to human oversight as a safeguard. But placing a person at the end of the workflow may not always be enough.
For instance: A manager who clicks “approve” without reviewing the facts is not providing meaningful oversight. Meaningful human review requires both information and authority, with a reviewer who does the following consistently:
Question the system’s result
Review the underlying information
Identify missing or inaccurate data
Consider protected leave, disability, or accommodations
Request more information
Override the recommendation
Document the final reasoning
The person should independently evaluate the individual in question, rather than simply confirm what the software suggested.
Vendor Technology Does Not Remove Employer Responsibility
Employers should not assume that a vendor carries all responsibility for AI hiring compliance.
According to the Equal Employment Opportunity Commission, federal employment discrimination laws apply when employers use AI and other automated systems, just as they apply to other employment practices, regardless of whether the tool was from a third-party vendor or not.
Put another way, while the vendor may provide the tool, the employer often decides which jobs the tool evaluates, what information enters the system, how scores are interpreted, who views the results, and whether candidates are screened out.
Employers remain accountable for how the tool is used.
Is It Illegal to Use AI in Hiring?
No. AI is not categorically illegal in hiring.
However, existing federal anti-discrimination laws around the use of AI still apply. Some state and local laws have also created specific rules for automated employment decision tools.
Depending on the location and use, employers may face additional requirements involving the following local regulations:
Candidate or employee notices
Bias audits
Impact assessments
Data disclosures
Record retention
Risk-management policies
Accommodation procedures
Human review
Appeal or correction rights
How Employers Can Reduce AI Hiring Risk
A seven-step checklist to review before adopting, configuring, or relying on AI-enabled hiring
1. Inventory Every AI-Enabled Tool
Create a list of the technology used for recruiting, screening, interviewing, scheduling, performance management, promotion, layoffs, and workforce analytics.
Do not limit the review to products marketed as AI. Ask existing software vendors whether automated scoring, ranking, machine learning, or recommendation features are built into their platforms.
2. Identify Which Tools Influence Important Decisions
Determine whether each tool does the following tasks, and document them appropriately:
Handles an administrative task
Provides information to a reviewer
Recommends an outcome
Ranks individuals
Screens candidates out
Substantially influences the final decision
Makes a fully automated decision
A scheduling assistant and an automated rejection tool should not receive the same level of oversight.
3. Review the Data Inputs
For each input, ask the following questions:
Is the information related to the job?
Is it accurate and current?
Could protected leave affect it?
Could a disability or accommodation affect it?
Could it act as a proxy for a protected characteristic?
Can the person correct inaccurate information?
Is the system missing important context?
Even a well-designed model can produce an unreliable result when it receives incomplete or misleading data.
4. Require Vendor Transparency
Employers should ask vendors to explain the following elements of their tools, to avoid a black-box issue:
What the system evaluates
What data it collects
How results are generated
How the model has been tested
Whether it has been tested for adverse impact
How often it is audited
How updates are documented
Whether results can be challenged
What information human reviewers can see
Whether the employer can override a recommendation
If a vendor cannot answer basic questions about a tool that influences employment decisions, consider looking for alternatives.
5. Test for Unequal Outcomes
Employers should not rely only on a vendor’s statement that a tool is “fair” or “unbiased.”
Manually review all outcomes and data in context of the actual jobs, candidates, locations, and uses of the system.
6. Preserve Meaningful Human Oversight
Assign trained people to review recommendations before any important decisions are made.
These reviewers should understand the purpose of the tool, its limits, the information it uses, and the situations in which it may produce an unreliable result.
Most importantly, they must have genuine authority to disagree with the system.
7. Create an Accommodation and Appeal Process
Candidates and employees should have a clear way to take the following steps:
Request a reasonable accommodation
Report an error
Correct inaccurate information
Provide missing context
Ask for human review
Challenge a result when appropriate
The process should be accessible, easy to find, and handled by people who understand the employer’s responsibilities.
AI Can Support Hiring, but Employers Remain Accountable
AI in hiring can improve speed, consistency, and organization; however, those benefits do not remove an employer’s legal or ethical responsibilities. Every hiring team remains responsible for how a tool is selected, configured, tested, monitored, and used.
AI should support responsible hiring. It should never replace accountable judgment.
Frequently Asked Questions About AI in Hiring
Yes. AI is not generally prohibited in hiring. However, employers must still follow federal, state, and local employment laws. Depending on the tool and location, employers may also need to provide notices, conduct a bias audit, document how the system is used, or offer meaningful human review.
The main risks include discrimination, inaccurate data, limited transparency, weak human oversight, and failure to follow applicable AI hiring laws. Employers may also face risk when they rely on a vendor’s recommendation without understanding how the result was produced.
AI can produce unequal results when it relies on biased historical data, incomplete information, or factors that affect certain groups differently. For example, a system may misread employment gaps, disability-related test results, protected leave, or schedule limitations as signs that someone is less qualified.
Algorithmic bias occurs when an automated system produces unfair or unequal outcomes. The bias may come from the data used to build the system, the factors selected for review, the way those factors are weighted, or how the employer uses the final recommendation.
Yes. Using a third-party vendor does not automatically remove the employer’s responsibility. Employers should understand what the tool evaluates, how it influences employment decisions, and what safeguards are in place before relying on its results.
A tool may be described as a black box when it produces a score or recommendation without clearly explaining how it reached that result. Not every AI system is unexplainable, but limited vendor information or poor documentation can make a result difficult to review, challenge, or defend.
No. Human review can reduce risk, but only when it is meaningful. A reviewer should have access to the relevant information and the authority to question, correct, or override the system. Simply approving an automated recommendation is not enough.
Meaningful human oversight means that a trained person independently reviews the facts behind a recommendation. The reviewer should be able to identify errors, consider accommodations or protected leave, request more information, and document why the final decision was made.
Employers should ask what data the tool uses, how results are created, whether the system has been tested for adverse impact, how often it is audited, and whether candidates can challenge errors. Employers should also confirm that human reviewers can see the relevant information and override the tool’s recommendation.
AI may be used to organize information, route reports, identify possible data issues, or support other parts of a screening workflow. However, background-screening information should support an individualized and job-related decision rather than cause an unexplained automatic rejection.
Employers can reduce risk by inventorying their AI-enabled tools, reviewing the data each system uses, testing for unequal outcomes, requiring vendor transparency, documenting governance, and preserving meaningful human oversight. They should also provide clear processes for accommodations, corrections, and appeals.
The plaintiffs allege that AI-assisted performance measures and activity data disadvantaged employees who took certain forms of leave or used accommodations. The company disputes the claims and has stated that people, rather than AI, made the workforce decisions. The allegations have not been proven.
How DISA Can Help
DISA Global Solutions helps organizations build structured screening programs that combine efficient technology with reliable information and human decision-making.
DISA Global Solutions aims to provide accurate and informative content for educational purposes only and does not constitute legal advice. The reader retains full responsibility for the use of the information contained herein. Always consult with a professional or legal expert.
Eden Hutchinson
Compliance Investigation Manager
DISA Global Solutions
Eden has a strong passion for quality, compliance, and background screening.
Learn the key differences between instant and comprehensive background checks to make informed hiring decisions. Discover which method ensures accuracy, complia...
The U.S. Department of Homeland Security (DHS) has announced that the COVID-19 temporary flexibilities for Form I-9, Employment Eligibility Verification will en...
Learn how to navigate the FCRA adverse action process to ensure compliant hiring practices and reduce legal risks for your business. Learn best practices....
Instant background checks promise speed, but staffing firms often pay later through errors, duplicate screening, and compliance risk. Learn how to balance speed...
August 21, 2026
11
min read
