# AI in hiring regulations: a general overview for recruiters

> A plain overview of AI in hiring regulations: what the rules usually require, which tools are covered, what to ask vendors, and a routine you can run.

Source: https://deephire.com/blog/ai-in-hiring-regulations-overview
Last checked against the product: 2026-11-17

## Key takeaways

- Existing anti-discrimination law still applies when AI is part of a hiring decision, even where no AI-specific law exists.
- Most AI hiring rules ask for the same few things: notice, consent in some places, bias testing, human review, and limits on data.
- List every tool that scores or filters candidates, and note where your candidates and jobs are located.
- Ask vendors in writing for their bias testing results and data retention terms. Also ask how a recruiter can override the tool.
- Rules differ from one place to another, so check with your legal team before you roll out a new tool.

AI in hiring regulations are rules that control how employers use software to screen, score, rank, or interview candidates. Most of them ask for the same few things: tell candidates when AI is involved, test the tool for bias, keep a human in the decision, and limit what you do with candidate data.

The details change from place to place, and new rules keep arriving. This guide covers the general picture and the most common requirements, then gives you a routine you can run without being a lawyer. It is general guidance only. Rules vary by location, so check with your legal team before you change your process.

## What counts as AI in hiring under these rules?

Most rules don't care what the vendor calls the product. They care what the tool does to a candidate's chances. A tool is usually in scope if it helps decide who moves forward. That includes software that:

- Scores or ranks resumes against a job description
- Filters applicants out based on keyword matches or test results
- Analyzes video or audio interviews and produces a score or recommendation
- Writes summaries that a recruiter relies on to advance or reject people
- Chats with candidates and screens them out based on their replies

Tools that only schedule interviews or send reminders are less likely to be covered, but that depends on how each rule is worded. A useful test is whether the tool's output could be the reason a person did not get the job. If it could, treat it as covered until your legal team tells you otherwise.

## Which laws apply to AI hiring tools?

Usually three layers apply at once. The newest AI laws get the headlines, but the older anti-discrimination and privacy laws apply to every hiring decision, so don't skip them.

### Anti-discrimination law still covers AI decisions

In the United States, federal anti-discrimination laws apply to hiring decisions no matter how they are made. If an AI tool screens out people of a certain race, sex, age group, or disability status at a higher rate, the employer can be held responsible. Pointing to the vendor generally does not move that responsibility off the employer.

Disability is where recorded and timed steps often go wrong. A timed assessment or a video analysis tool can put candidates who need accommodations at a disadvantage. See [ADA hiring basics for recruiters](https://deephire.com/blog/ada-basics-for-hiring-process) for how accommodation requests work during interviews.

### Rules written for AI and automated tools

Some local and national governments have passed rules aimed at automated hiring tools. You will often hear about these:

- New York City's rule on automated employment decision tools, which requires bias audits and candidate notice
- Illinois's law on AI analysis of video interviews, which requires employers to explain the analysis to candidates and get their consent
- Colorado's broader AI law, which covers employment decisions
- The EU AI Act, which treats AI used for hiring as high risk and puts duties on both vendors and employers

These rules define a covered tool in different ways. They also disagree on who must run the audit and what the notice must say. Some have been amended or delayed after passing. Ask your legal team which ones apply where your candidates live and where your jobs are based.

### Privacy and data laws

Privacy laws control how you handle candidate data from collection through deletion. Some states have rules on biometric data, which can include face geometry or voiceprints. Some places give candidates the right to ask what data you hold or to have it deleted. If an AI tool processes recordings or facial data, privacy rules may apply even where no AI-specific law exists.

## What do most AI hiring regulations require?

The wording varies, but the same requirements keep showing up. This table turns them into actions.

| Requirement | What it usually means | What you can do today |
|---|---|---|
| Notice | Tell candidates AI is used and what it evaluates | Add a plain sentence to job posts and interview invites |
| Consent | Get agreement before AI analyzes the candidate, in some places | Add a checkbox or written consent step before the interview |
| Bias testing | Check whether outcomes differ by group | Run or request a bias audit at least once a year |
| Human review | A person makes or can change the final decision | Document who reviews AI output and when |
| Alternative process | Offer another way to be evaluated, in some places | Have a phone or live interview option ready |
| Data limits | Limit who can see candidate data and how long you keep it | Set a retention period and a deletion request process |
| Explanation | Describe how the tool works in general terms | Keep a short plain description from your vendor on file |

Bias testing is the row most teams find hardest. For what an audit measures and how to read the results, use the walkthrough on [running an AI hiring bias audit](https://deephire.com/blog/ai-hiring-bias-audits).

## How to check whether your tools are covered

Start with an accurate picture of your own process. Work through these steps with your team.

1. List every tool in your hiring process, from the job board to the offer letter.
2. Mark each tool that scores, ranks, filters, summarizes, or recommends candidates.
3. For each marked tool, write down whether a human sees every candidate or only the ones the tool passes through.
4. Note where your candidates are located and where the jobs are based. Remote roles can pull in rules from many places.
5. Collect the vendor's documentation on how the tool works, including what data it uses and how long it keeps that data.
6. Bring this list to your legal team and ask which rules apply to each tool.

Spend the most time on step 3. A tool that auto-rejects candidates nobody ever looks at carries far more risk than one that offers a recruiter advice on someone they are already reviewing.

## What should you ask an AI hiring vendor?

A vendor should be able to answer these questions in writing. If they can't, treat that as a warning sign.

- What exactly does the AI evaluate, and what does it ignore?
- Has the tool been tested for bias, by whom, and can we see the results?
- Can we set our own criteria, or does the tool use a fixed model?
- Can a recruiter override or ignore every AI output?
- Does the tool use biometric data, such as facial or voice analysis?
- Where is candidate data stored, and how long is it kept?
- How do we handle a candidate's request to delete their data?
- What notice and consent features are built in?
- Will you tell us when the model or scoring method changes?

Keep the answers in your vendor file. A bias audit or legal review will usually ask for exactly this information.

## Where do one-way video interviews fit?

A one-way video interview, where candidates record answers on their own time, is not AI by itself. It becomes an AI question when software analyzes the recordings and produces scores or summaries. Some rules focus on this exact case, so disclosure and consent matter here.

The lower-risk setup is one where AI gives advice and a human decides. In DeepHire, for example, AI summaries and scores are based on the recruiter's own rubric and are advice only. The recruiter still watches the answers and makes the call. Whatever tool you use, tell candidates up front what happens to their recordings. You can adapt the sample wording in [AI video interview disclosure](https://deephire.com/blog/ai-video-interview-disclosure).

A notice line for an interview invite might read:

> We use software to help summarize your recorded answers. A recruiter reviews every interview and makes all decisions.

Only use a line like this if it is true. If recruiters skip low-scored interviews, the notice is promising candidates something you don't do. Change the process or change the wording.

## Common mistakes recruiters make with AI hiring tools

Most of these come from nobody owning the tool's settings.

- Turning on AI features by default without telling anyone on the team.
- Letting a score become the decision, so nobody watches or reads the low scorers.
- Using the same notice text everywhere, even where consent is required.
- Keeping recordings and resumes forever because nobody set a retention period.
- Assuming the vendor handles compliance. Under most rules, the employer has duties too.
- Forgetting accommodation requests for timed or recorded steps.
- Missing new AI features that arrive in a vendor update.

A written policy with a named owner prevents most of these. Start from the [AI hiring policy template](https://deephire.com/blog/ai-hiring-policy-template) and adjust it with your legal team.

## A simple routine to stay current

Because the rules keep changing, a short recurring checklist works better than a one-time cleanup.

### Monthly checks on tools and data requests

Check whether any vendor has added AI features. Review a sample of candidates the AI rated low and confirm a human looked at them. Handle any data deletion requests.

### Quarterly review of notices and new rules

Compare your notice and consent wording against what your tools actually do. Ask your legal team whether any new rules affect the places you hire. If you collect demographic data, check pass-through rates at each stage for unexpected drops in any group.

### Yearly bias audit and policy update

Run or request a bias audit for each covered tool and update your AI hiring policy to match. Then retrain recruiters so they treat AI output as advice, not as the final answer.

Keep a simple log of what you checked and when. When someone outside the team asks how you use AI, you can show them the log instead of reconstructing it from memory.

## Frequently asked questions

### Is it legal to use AI to screen job applicants?

In most places, yes, but the same anti-discrimination laws that cover human decisions still apply. Some places add their own requirements, such as candidate notice or bias audits. Check with your legal team about the places where you hire.

### Do I have to tell candidates I am using AI in hiring?

Several AI hiring rules require notice, and some also require consent before AI analyzes an interview. Where it is not required, telling candidates is still a sensible default. Use plain wording that matches what your tool actually does.

### Who is responsible if an AI hiring tool is biased, the employer or the vendor?

The employer usually stays responsible for its hiring decisions, even when a vendor built the tool. Some newer rules also place duties on vendors. Get bias testing results from your vendor and keep a human in every decision.

### Does a one-way video interview count as AI under hiring laws?

Recording answers is not AI by itself. It can fall under AI rules when software analyzes the video or audio and produces scores or recommendations. If yours does, review your consent and data handling with your legal team.

### How often should AI hiring tools be audited for bias?

Some rules require an audit at least once a year, and that is a sensible baseline even where it is not required. Also review a tool when the vendor changes the model or when you start using it for a new type of role.
