How to Avoid Algorithmic Bias and AI Risk in Hiring

Generations of people have joined and left the workforce since U.S. employment discrimination laws took effect in the 1960s, but problems continue to surface in old and new ways. For example, many companies now use artificial intelligence to support efficient hiring and minimize human bias. But it turns out that algorithms can learn bias, and that bias is not always easy to detect.

How do these developments affect compliance, reputation and litigation risk? And how can in-house legal teams guide procurement and human resources on vendor agreements, as well as use of AI-assisted recruiting tools?

Why In-House Guidance Is Important Now

Trend: Legislative, academic and regulatory interest in governance over AI-assisted hiring is on the rise.

Risk: Non-compliance with state and local measures, federal guidance and regulations concerning algorithmic bias creates new legal and reputational exposure.

Recommended In-House Actions:

• Understand how algorithmic bias can develop.
• Inventory the AI tools you use in HR and hiring.
• Vet AI tools to identify risks, including algorithmic bias.
• Maintain and regularly train employees on policies and procedures to manage risks.
• Stay aware of pending and new legislative measures, as well as litigation.

Specific examples below can help you plan your next steps. (6 minutes).

Growing Prevalence of AI-Assisted HR Tools

AI and automation are now widespread in recruiting: 87% of companies and 99% of Fortune 500 firms use AI tools to aid them in their hiring processes.(1) These tools can help in drafting position descriptions, screening resumes, evaluating recorded video interviews, assessing skills and more. But without an informed and disciplined approach, the desired efficiency can carry significant employment discrimination risk.

Beyond Business Leaders, Who Is Watching the Impact of AI in HR Decisions?

Lawmakers: AI in hiring has drawn attention from legislators considering legal frameworks for AI. Legislators in 46 states and in Congress have considered more than 1,700 AI-related bills so far during 2026; 39 states and the U.S. Congress have enacted AI legislation.(2)

Some bills could require employers to disclose AI-related or automation-related job displacement, layoffs or hiring disruption. Several states and New York City now have laws that address AI workplace discrimination, require employers to provide notice when using AI for employment decisions, or both.(3)

Connecticut recently enacted CT SB5, the Artificial Intelligence Responsibility and Transparency Act, which requires an inventory of state agency systems that use AI and impact assessments of those systems. It amends the Connecticut Fair Employment Practices Act to confirm that use of automated employment decision technology is not a defense to a discrimination claim, although a court may weigh an employer’s anti-bias testing in mitigation. Its provisions phase in on dates ranging from October 2026 through January 2028.(4)

Across federal, state and local activity, a primary focus is the possibility of algorithmic bias and the need to take affirmative steps to prevent, identify and rectify it.(5) For related insights, see the August 2026 State Net Capitol Journal™ article States Take Different Approaches to Regulate AI in Employment.

Academics: After analyzing 4 million job applications submitted for 1,700 postings across 150 employers and 11 industries, Stanford University researchers reported in May 2026 that an algorithm used to screen and rank applicants produced adverse impact against Black and Asian applicants; 26% of Black applicants and 15% of Asian applicants applied to at least one position where the tool disadvantaged their racial group. The tool returned “recommend” or “do not recommend” labels that informed, rather than replaced, employer hiring decisions. Researchers applied the “four-fifths rule” from the Uniform Guidelines on Employee Selection Procedures, adopted by the Equal Employment Opportunity Commission (EEOC) and other federal agencies. The rule flags a position when one group is recommended at less than 80% of the rate of the most-recommended group. The researchers found that looking at positions individually exposed adverse impact that aggregate averages could hide.(6)

Regulatory Professionals and Litigators: Employers should expect the EEOC to continue focusing on recruiting and hiring issues, and the private plaintiffs’ bar to follow suit.(7)

How Algorithmic Bias Can Develop in Your Decision Tools

Some companies use AI tools for drafting job descriptions in an effort to remove unconscious bias and attract a more inclusive pool of candidates.(8) Yet a tool can acquire biases that are not evident when it is first used. One way this happens is through training data.

Developers train many AI tools using machine learning algorithms that analyze vast amounts of historical data. Some tools continue to learn as they process more data about new applicants or existing employees. Bias is not always apparent in a dataset, and the data can inadvertently introduce bias into the processes and outcomes of a tool that initially appeared neutral.(7)

How can this happen?

• Training datasets could be unrepresentative, incorporate historical biases or include variables that correlate with protected characteristics.
• When training data comes largely from successful applicants who predominantly share a particular gender, ethnicity or other protected characteristic, the system may learn to favor that group.
• Algorithms may exclude candidates based on demographics or unfairly penalize certain language or speech patterns.
• Seemingly neutral factors—such as years of work experience—may disadvantage younger applicants or those who have taken breaks from work, like parental leave.

Gain Advantages While Reducing Risk

Here are ways to benefit from AI advantages and mitigate risk related to algorithmic bias.

• Inventory your AI hiring tools, classify them under applicable state frameworks, implement jurisdiction-specific notice and adverse-decision procedures, and maintain enforcement-ready governance records that can withstand both state regulatory scrutiny and federal investigation.
• Maintain clear documentation showing that AI hiring tools consistently apply criteria that are job related for the position and consistent with business necessity.
• Regularly review hiring and recruiting processes, including how you vet technologies.
• Require vendor contracts to include representations that the vendor has tested its automated systems for bias and that those systems comply with applicable anti-discrimination laws.
• Preserve the data you need to defend any AI-assisted decision and to demonstrate a legitimate, nondiscriminatory reason for each outcome.
• Ensure your process preserves human oversight in all referral and hiring decisions.

Tools to Assist Your In-House Team

The Protégé In-House™ solution from LexisNexis® enables you to generate surveys of laws and regulations and expedite contract drafting aligned with your company’s priorities, style and standards. Answers and summaries are grounded in Practical Guidance content and other authoritative sources. LexisNexis® State Net® provides legislative and regulatory analysis, reporting and AI summaries to help you track compliance obligations. The Nexis Diligence+™ solution connects you to intelligence for vetting customers, suppliers, contractors and other third parties—including your AI system providers.

Advancing a Proactive Approach

AI can make recruiting and hiring less time-consuming, but unintended bias may affect technology as well as people. The steps above can help you protect your company and advance HR initiatives to hire candidates who truly match your business needs.

Endnotes

1. DataRefs, AI Recruitment Statistics (2026) – Hiring Trends & Market Size.

2. LexisNexis® State Net® data, July 2026.

3. AI’s Impact on Workplace Remains Focus for State Legislators, State Net Capitol Journal, June 30, 2026.

4. Artificial Intelligence State Law Survey, Practical Guidance, July 23, 2026. See also Connecticut Artificial Intelligence Responsibility and Transparency Act, 2026 Conn. SB 5.

5. Artificial Intelligence in the Workplace: Best Practices, Practical Guidance, current as of July 15, 2026.

6. AI Hiring Tools Can Yield Racial Bias and Systemic Rejection, HAI Stanford University Human-Centered Artificial Intelligence, May 26, 2026. See also Uniform Guidelines on Employee Selection Procedures, 29 C.F.R. § 1607.4(D) (four-fifths rule).

7. Law360 Employment Authority, “Flashpoints In Focus: Tips As EEOC Prioritizes Hiring Bias,” May 15, 2026.

8. AI in Employment Decisions and Performance Management: Key Legal Issues and Potential Risks and Benefits, Practical Guidance, current as of July 31, 2026.