Employers don’t need to build AI algorithms to be held accountable.

Companies are rapidly incorporating artificial intelligence into workplace decision-making for fear of being left behind. Employers routinely use AI tools to screen applicants and resumes, analyze interviews, monitor employee productivity, and make promotion and disciplinary decisions.

While some companies and HR leaders assume that the legal risk lies with the software vendor who created or licensed the tool, this is not the case in many situations. As with other tools used by businesses, employers can be held liable for decisions made with AI assistance, even if the technology is purchased or licensed from a vendor.

Many states have increased oversight of AI-based employment decisions, and further state action is expected in the future. This is especially true given the current administration’s position regarding enforcement of disparate impact claims, including the rescission of certain AI guidance issued by the U.S. Equal Employment Opportunity Commission in the past several years.

Federal approach

Under the Biden administration, the EEOC issued guidance regarding the use of AI in employment decisions by employers. The guidance alerted employers and raised concerns about the potential for disparate impact discrimination, disability discrimination, lack of transparency in the use of AI tools, bias in training data, and lack of human oversight. It also clarified that employers may be held liable for discrimination that occurs through the use of tools developed or purchased from third parties, even if there is no intent to discriminate.

In early 2025, the Trump administration issued an executive order directing federal agencies to deprioritize disparate impact enforcement, and shortly thereafter the EEOC rescinded its guidance on AI.

While the current administration’s actions may indicate to employers that AI-related risks in the context of discrimination are no longer a concern, this is not the case. The doctrine of disparate impact discrimination, which focuses on outcomes rather than intent, remains valid under federal court precedent. Neither an executive order nor a revocation of agency guidance can overturn long-standing legal precedent.

State and local actions

In addition, states are actively stepping in to fill the gaps left by federal law. Illinois currently prohibits the use of AI that results in employment discrimination based on a protected characteristic, even if the discrimination was unintentional, and requires employers to disclose when AI tools are being used. New York City requires bias audits of AI tools used in hiring decisions and requires employers to notify candidates and employees about their use. Several other states have passed or proposed similar laws, codifying protections against disparate impacts or putting in place specific guardrails for how employers use AI in hiring and workforce decisions.

As such, EEOC enforcement for various effects may be reduced, but litigation risk will not be completely eliminated. State agencies and individuals can bring these claims against employers under both state and federal law.

What employers need to understand

Organizations may begin using AI for recruitment and employee decisions without knowing what data the system relies on, how the AI ​​was trained, whether protected characteristics are indirectly taken into account, whether the results are negative, and how to account for the decision-making process.

This lack of understanding creates risk, and in the event of a lawsuit, courts are unlikely to trust defense attorneys who blame the creator or vendor.

HR teams need a clear understanding of where AI tools are being used across the organization. Many employers use tools, sometimes without even realizing it, in the decision-making process for screening applicants, interviewing, scheduling, monitoring productivity, and promoting and disciplinary decisions.

Employers also need to understand how the tools are being used in each context. Are managers implementing AI-suggested outcomes without reviewing them? Or is there human oversight and review involved? Relying on AI to make decisions without sufficient human oversight increases risk.

protect against discrimination claims

After gaining a better understanding, employers should consider whether these tools are producing different results. Bias audits may be required under state or local law, but even if they are not, they are one of the strongest defenses available in the event of a claim. Employers should assess whether there is documentation that explains what audits and controls are in place to ensure the tool is not producing these disparate results.

You will also need documentation that explains how AI is being used to make decisions in the workplace, including internal policies, vendor materials, and training documents regarding employee use of AI and human oversight.

Employers should also ensure that the use of AI tools is disclosed as required under state and local law and, if necessary, explain to employees and applicants how the tools will be used.

Finally, employers should review their vendor contracts to understand what promises the vendor makes about how the tools are trained, what audits are conducted, and who will be held accountable. If your contracts are not aligned with your organization’s compliance obligations, it’s best to get on the same page about those policies now, rather than after a complaint has been filed.

Employers that can demonstrate a compliant implementation are best placed to defend claims based on the employer’s use of AI tools. Technology will continue to evolve faster than the laws that govern it, but the fundamental responsibilities remain the same.

Employers are responsible for the consequences of their decisions in the workplace, whether those decisions are made by humans, algorithms, or something in between.