What’s actually slowing this PC down?
Pick the symptom - the matching free tool is one click away.
Using an AI tool to screen, score, or rank job candidates does not move legal responsibility to the software or the company that sold it. In the United States, the employer that makes or informs a selection decision remains answerable under civil-rights law. A human manager in the loop is a possible safeguard, but only if that review is shown to be meaningful. The tool can make decisions more consistent and scalable, and it can also reproduce bias, exclude disabled applicants, or make the reasoning harder to inspect.
Who is accountable when AI makes a hiring decision?
The employer is. The U.S. Equal Employment Opportunity Commission (EEOC) has stated that Title VII of the Civil Rights Act applies to automated systems that employers use to make or inform selection decisions. The agency summarized that position in its 2023 annual performance report, and it covers recruitment, hiring, monitoring, and firing. A score, ranking, or recommendation supplied by a third-party product therefore does not shift the question of responsibility away from the employer that relies on it.
The EEOC framed the issue in its October 28, 2021 announcement of its AI and Algorithmic Fairness Initiative. Chair Charlotte A. Burrows said: “While the technology may be evolving, anti-discrimination laws still apply.” In the same announcement she added that the agency “will address workplace bias that violates federal civil rights laws regardless of the form it takes, and the agency is committed to helping employers understand how to benefit from these new technologies while also complying with employment laws.” These are official agency statements, not a court holding, but they set out the agency’s enforcement position.
Can an employer blame a hiring algorithm?
Not as a defense. New York City Commission on Human Rights guidance on disability discrimination says covered entities are responsible for the actions and decision-making of AI systems and other technology they use. It also says an employer may not avoid liability for unlawful discrimination by claiming the technology caused it. That is city guidance, not federal law, but it states the principle plainly: a vendor’s product, configuration, or recommendation does not replace the employer’s own obligations.
Recommended Free Tools
#1 Best Overall
Practically, this means an employer should be able to answer four questions about any automated step: what the tool measures, why those measures relate to the job, who reviewed its output, and how a rejected candidate could raise a problem with the result.
Does human review make an AI hiring decision fair?
Not automatically. Human managers are not a clean control group. Official guidance does not establish that human judgment is unbiased, that AI is more biased than managers in every setting, or that placing a person somewhere in the workflow prevents liability. A manager who rubber-stamps a score has added a signature, not a safeguard, and a manager’s own subjective impressions can produce discriminatory outcomes just as an automated process can.
The comparison is therefore about process quality, not about who makes the call. The table below sets out the practical axes on which the two approaches differ.
| Axis | AI-supported process | Human-manager process |
|---|---|---|
| Consistency | Can apply the same stated process across many records. Consistency does not establish validity or fairness. | Judgment can vary by reviewer and context. Official guidance cited here does not quantify how much. |
| Evidence | May produce scores or rankings that need explanation, validation, and impact review. | May rely on interviews, references, or impressions. Job-related grounds should be documented. |
| Bias and access | Can reproduce patterns in training or historical data, or disadvantage disabled people through test or interface design. | Can also produce discriminatory outcomes. Human judgment is not automatically safe. |
| Accountability | The employer remains subject to applicable obligations. Vendor involvement does not remove them. | The employer remains accountable for its own decision and process. |
| Challenge and correction | Requires notice where the law demands it, a route to an accommodation or alternative process, and a real way to correct errors. | Requires an identifiable decision-maker and a record of the reasons and evidence considered. |
A human review adds value only when it meets a defined standard. In practice, that means the reviewer:
Rank #3
- has the authority to override or question a tool’s recommendation, not just to approve it;
- receives the job-related information needed to evaluate the recommendation, including what the tool measured;
- considers accommodation requests and disability-related issues before the final action;
- records the specific reasons for the final decision; and
- is identifiable, so a candidate or regulator can ask who made the decision and on what basis.
These are governance criteria drawn from the agencies’ concerns, not a quoted legal test. Whether a particular process satisfies a particular statute should be checked against that statute.
Where the risks concentrate
Bias that is amplified, new, or hidden
The New York State Office of the State Comptroller’s 2025 audit of New York City’s automated employment decision tool enforcement describes tools that scan resumes, analyze online presence, and evaluate video interviews. The audit identifies three risks: tools can amplify existing bias, they can introduce new sources of bias, and vendors often give weak information about what a tool can and cannot do. The last point matters because an employer cannot test for limits it has not been told about.
Disability access
The EEOC and the Department of Justice warn that automated tests and software can screen out a person with a disability who could do the job with or without reasonable accommodation. Tools can also prompt disability-related inquiries that the Americans with Disabilities Act restricts. Employers should check whether a timed test, a video format, or an interface that requires particular motor or sensory abilities creates a barrier, and should give candidates a way to ask for a different process.
Disparate impact and the four-fifths rule
The EEOC’s Title VII guidance, as summarized in its 2023 annual performance report, says employers should assess whether an automated selection procedure creates disparate impact on protected groups. It also cautions that meeting the four-fifths rule in the Uniform Guidelines does not guarantee that a procedure will be found free of prohibited disparate impact. A passing ratio is a screening measure, not proof of fairness, and an employer should not treat it as one.
What New York City Local Law 144 requires
Local Law 144 has a narrower scope than the federal guidance. It covers an automated employment decision tool used to screen a candidate or employee for an employment decision, and it applies to covered employers and employment agencies in New York City. An employer that uses workplace software for other purposes is not automatically covered. For covered use, the Administrative Code sets these steps:
- Obtain a bias audit within one year before use. The audit must be conducted no more than one year before the tool is used.
- Publish the audit summary before use. The most recent audit summary and the distribution date of the tool version audited must be publicly available before the tool is used.
- Give notice at least ten business days before use. The notice must state that the tool is being used and identify the qualifications or characteristics it assesses.
- Offer an alternative. The notice must allow a candidate to request an alternative selection process or an accommodation.
- Disclose data practices on request. If the data types collected, the data sources, and the retention policy are not on the employer’s website, they must be provided within 30 days of a written request.
These requirements apply only in New York City. They should not be described as national law.
What the enforcement record shows
The Comptroller’s audit reviewed the New York City Department of Consumer and Worker Protection (DCWP) for July 2023 through June 2025. The figures below are the ones the audit reports, and each carries its own qualification.
- One potential compliance issue. DCWP’s review of 32 company websites and bias audits identified one issue. The Comptroller’s review of the same companies identified at least 17 potential instances. Both are potential non-compliance, not adjudicated violations.
- Two complaints received. DCWP recorded two complaints about automated employment decision tools during the audit period. The Comptroller also found that DCWP did not investigate whether its complaint intake process was working.
The audit’s broader finding is that enforcement depends on people noticing violations. DCWP told the Comptroller that complaint-based enforcement is difficult because an organization that decides it is outside the law may not post audits or notices, which makes violations harder to identify. The Comptroller judged the complaint routing process ineffective. A published audit summary is therefore a useful signal, but its absence does not prove compliance.
Limits of this guidance
- Jurisdiction. The federal material above concerns U.S. civil-rights law. State, local, and international requirements differ and are not covered here.
- Currency. The EEOC statements cited date from October 2021 and its 2023 annual performance report. Agency guidance is revised or withdrawn over time, so confirm each document’s current status before relying on it.
- Voluntary frameworks. The National Institute of Standards and Technology’s AI Risk Management Framework 1.0, released January 26, 2023, is voluntary guidance intended to help organizations build trustworthiness into AI systems. NIST says the framework is being revised. It is not an employment law and it does not replace legal advice.
- Code currency. The New York City Administrative Code as published online may not reflect the latest legislation or rules, so check the current legal text and the DCWP guidance before acting on notice details.
Employers facing a specific hiring tool should treat this article as a map of obligations and risks, and should take legal advice on whether a particular process meets them.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
