You’re hiring for several positions. You get 105 résumés for a single role and you don’t have time to read them all. You mention your situation to a colleague and they say: “use ChatGPT” or “there’s an AI tool that screens résumés for you.” It sounds perfect. You save hours, cut down on administrative tasks, get to the candidates who actually fit the role faster, and you can return to your “real” work more quickly.
However, there is an important question almost nobody asks before uploading résumés to an AI tool for screening: who’s responsible if the AI discriminatorily screens out candidates, even if you never intended it to discriminate?
The short answer: you are. The employer. Not the company that sold you the software.
Why Isn’t This Just an AI Problem?
U.S. employment and discrimination laws — Title VII of the Civil Rights Act, the Age Discrimination in Employment Act (ADEA), the Americans with Disabilities Act (ADA) — weren’t written with algorithms in mind, but they don’t need to be updated to apply. What matters under these laws isn’t how a decision to screen out a candidate was made, but whether the outcome had a disparate impact on a protected category, which can include sex, race, age, national origin, or disability, among others.
AI tools known as “Automated Employment Decision Tools” (“AEDTs”) can filter résumés by keyword, score video interviews by analyzing speech patterns and facial expressions, and rank candidates, among other tasks. The problem is that these systems are built by “training” them on historical hiring data, and if that data reflects past biases — for example, fewer women hired into engineering roles — the tool can replicate that bias at scale, quietly, with no one noticing it until a complaint arrives.
New York City was one of the first jurisdictions to regulate this. Its AEDT law (Local Law 144) requires that, before using one of these tools, an employer must conduct an independent bias audit, publish the results, and notify candidates that AI is being used in the process — and giving them the option to request an alternative process. That audit must calculate selection rates and “impact ratios” by sex, race/ethnicity, and intersectional categories. The law applies even if the AI is only used at an early stage of the process, such as scheduling interviews, and not for the final hiring decision.
Illinois has its own law (the AI Video Interview Act), which requires notifying candidates when AI is used to analyze video interviews. Colorado, under its privacy law rules, requires risk assessments and an opt-out right when an “AEDT” is used with legal or significant effects on a person. California is moving in the same direction with its regulations under CCPA.
As of this publication, Puerto Rico does not have a specific law on bias audits for AI hiring tools. That doesn’t relieve you of responsibility, though. On one hand, federal law still applies, and the Equal Employment Opportunity Commission (EEOC) has made clear that employers are responsible for the discriminatory impact of the tools they use, regardless of who programmed them. On the other hand, our own Law No. 100 of June 30, 1959, which prohibits discrimination in employment, applies here too, even though it doesn’t contain the word “algorithm.” In our next article, we’ll discuss in detail how a violation of Law 100 could bring serious consequences.
Why Should You Care?
Because liability for discrimination generated by AI is yours, not the software vendor’s. If you buy or subscribe to an AI tool to filter résumés and that tool ends up disproportionately screening out candidates over 40, women, or people with names the system (incorrectly) associates with lower English proficiency, you’ll be the employer sued — not the software company.
Think about it from the practical standpoint of a small business or a clinic in Puerto Rico:
- You don’t have a large legal department to review these software contracts.
- You probably don’t know what data the tool was trained on.
- It’s easy to assume that “if a big company sells it, it must be legal to use this way, no further questions asked.”
That last point is the most common mistake. No AI tool comes with a guarantee of legal compliance for your jurisdiction and your industry. The responsibility for making sure your hiring process doesn’t discriminate — whether AI is involved or not — is still yours.
There’s also an additional practical problem: if a rejected candidate asks why they weren’t selected, or files a complaint, can you explain how the process worked? If your only answer is “ChatGPT decided who we should select,” that’s not a legal defense. In fact, it’s evidence that there was no adequate human oversight.
How Can I Comply with the Law?
This doesn’t mean you should avoid AI in your hiring process. It means you need to use it with the same controls you’d apply to any important employment decision, with some of them being:
- Investigate the tool before using it. Ask the vendor whether an independent bias audit has been done, when the most recent one was, and whether they’re willing to share those results with your candidates if asked.
- Never let the AI have the final word. Use the tool to organize or prioritize, not to automatically reject candidates without human review. A human being should be able to explain and justify every decision.
- Notify candidates. Even though Puerto Rico doesn’t yet require it by law, it’s good practice — and a legal defense — to disclose in the job posting that AI technology will be used as part of the evaluation process.
- Document the process. Keep records of which tool you used, what criteria the system applied, and what human review took place before each rejection. If no one ever asks, you’ve lost nothing. If they do ask, that documentation is your defense.
- Have an alternative process. If a candidate doesn’t want their application evaluated by AI, set up a process where a person reviews their résumé directly.
- Review your vendor contract. Who’s responsible if the tool produces a discriminatory outcome? Do you have the right to audit how it works? These clauses are rarely negotiated, but you should try.
The Bottom Line
AI can genuinely save you time in the hiring process, and there’s nothing wrong with using it. The problem isn’t the technology — it’s treating it as if it were neutral, infallible, or someone else’s responsibility. If an AI tool screens out candidates in discriminatory ways, the law holds you, the employer, responsible. Use it as a support tool, keep human oversight on every decision, document your process, and verify — before you buy, not after a complaint — what controls the tool you’re using actually has.
Are you using or considering using AI in your hiring process and aren’t sure whether your current process exposes you to legal risk? Let’s talk. Schedule a consultation here.
