Canonical: [https://getkira.com/blog/ai-interview-laws-for-employers](https://getkira.com/blog/ai-interview-laws-for-employers)

↑Back to top

[←All posts](https://getkira.com/blog)

# AI Interview Laws: What Employers Need Before Turning On an AI Screening Tool

Vladimir Terekhov

Founder, Kira-AI

October 19, 2026 · 12 min read

![Abstract balance scale, shield, checklist, and connected nodes representing AI interview law compliance](https://getkira.com/_next/image?url=https%3A%2F%2Fcdn.sanity.io%2Fimages%2F44aj5pkb%2Fproduction%2F6e329b62e867241545ef8b9f53f86bc90807da33-2624x1632.webp&w=3840&q=75)

AI interview laws don't care whether your vendor calls a feature "AI." They care about where the job and the candidate are, what the tool actually does, what data it uses, and how much it shapes who moves forward. This guide gives you a fast way to classify a tool, a comparison of the rules in Illinois, New York City, Colorado, and the EU, and a five-file compliance packet to build before launch.

This is an operational guide, not legal advice. Artificial intelligence employment law moves quickly, and other states and countries have their own rules, so have counsel confirm coverage for your locations and exact workflow.

## Start With What the Tool Does, Not What It's Called

"Is this an AI tool?" gets a yes for almost everything in a modern ATS, so it's the wrong first question. The better one is how far the tool's output reaches into the hiring decision. Most AI interview laws scale their duties to that reach, under different names.

Use this ladder to classify each workflow:

| Level | What the tool does | Example | What it usually adds |
| --- | --- | --- | --- |
| Administrative | Moves work along without judging people | Scheduling, reminders, transcription | Notice where required, basic records |
| Advisory | Produces information a human reads | Answer summaries, flagged skills | Notice, checks on what gets surfaced, a documented human decision |
| Ranking | Scores or orders candidates | Fit scores, stack ranking | Bias testing or audit, stronger notice, review rules |
| Dispositive | Decides who advances or gets rejected | Auto-reject below a cutoff | Full evidence file, human review route, adverse-outcome handling |

The decision rule that holds across every jurisdiction below: **the more a tool shapes who advances, the more evidence, notice, and human review you need.**

Two traps live here. The level belongs to the workflow, not the product. And "administrative" doesn't mean exempt. Illinois's broader employment AI rule and the EU AI Act can both reach tools that look harmless on a feature sheet. If you're still deciding where AI belongs in your funnel, the [AI candidate screening guide](https://getkira.com/blog/ai-candidate-screening-guide) walks through the stages.

### One tool, two workflows

Picture two teams hiring warehouse supervisors with the same one-way interview platform and six questions.

Team A uses the transcript and an AI summary. A recruiter reads every summary, plays the recording when an answer looks thin, and decides who gets a call with the hiring manager.

Team B turns on the platform's 0-100 score and auto-rejects anyone under 65. Nobody looks at the rejected candidates.

Team B carries far more exposure. The score is now the only factor for a large share of applicants, which is close to the textbook case for New York City's automated employment decision tool rules and Colorado's "materially influence" standard, and it's the kind of candidate evaluation the EU AI Act treats as high-risk. Every rejection is an adverse outcome with no person involved. If the cutoff screens out a protected group at a higher rate, existing anti-discrimination law applies whether or not an AI-specific statute does.

Team A isn't automatically unregulated. Illinois requires notice when AI is used for recruitment, a summary can drop or soften a relevant answer, and an EU deployment that evaluates candidates can still be high-risk. Team A just has an easier time proving a person made the call. This breakdown of [automated interview evaluation](https://getkira.com/blog/automated-interview-evaluation) explains how scoring models turn answers into numbers, the part your notices need to describe.

## AI Interview Laws Compared: Illinois, NYC, Colorado, EU

Each row gets more detail below.

| Jurisdiction | Trigger | Main duties | Operator trap |
| --- | --- | --- | --- |
| Illinois: Video Interview Act | AI analysis of recorded video interviews for Illinois-based roles | Notice, explanation, and consent before the interview; sharing limits; deletion on request; demographic reporting if AI alone picks who gets in-person interviews | It only covers video, so teams wrongly assume Illinois is done |
| Illinois: Human Rights Act | AI used in recruitment, hiring, and other employment decisions | No discriminatory effect, no ZIP codes as a proxy, notice of AI use | Applies to audio and text tools the Video Act misses |
| New York City: Local Law 144 | Automated employment decision tool used "in the city" for hiring or promotion | Independent bias audit within the past year, published summary, notice 10 business days before use | A vendor's audit doesn't move the duty off the employer |
| Colorado: ADMT Act | Technology that materially influences employment decisions | Notice, post-adverse-outcome disclosure, data access and correction, meaningful human review | Summaries of the original act are out of date |
| EU: AI Act | AI intended for recruitment, selection, or candidate evaluation | High-risk controls, human oversight, logs, informing candidates; workplace emotion recognition banned | Duties phase in; can reach non-EU employers hiring for EU-based roles |

## Illinois: Two Laws With Different Triggers

### The Artificial Intelligence Video Interview Act

The [Illinois Artificial Intelligence Video Interview Act](https://www.ilga.gov/Legislation/ILCS/Articles?ActID=4015&ChapterID=68&Print=True) applies when you ask an applicant to record a video interview and use AI to analyze it for a position based in Illinois. Before the interview, you must tell the applicant that AI may be used, explain generally how it works and what types of characteristics it evaluates, and get their consent.

You can't use AI to evaluate anyone who doesn't consent. Videos can only go to people whose expertise or technology is needed to evaluate the applicant. If a candidate asks, you must delete their interview and tell everyone who received a copy to delete theirs, backups included, within 30 days.

If you rely solely on AI analysis of the video to decide who gets an in-person interview, you also have to collect and report demographic data on who was and wasn't advanced.

### The Human Rights Act amendment

Illinois separately amended its Human Rights Act, and that change is now in effect. It makes it a civil-rights violation to use AI in recruitment and other employment decisions in a way that discriminates against protected classes. It also bans using ZIP codes as a proxy for those classes and requires notice when AI is used for covered employment purposes. Notice details depend on the statute and implementing rules, so check the current version rather than copying another company's template.

The trap: teams running audio-only or text interviews conclude Illinois doesn't apply because they don't record video. The Video Act may not reach them. The Human Rights Act amendment still does.

## New York City: The Best-Known AI Hiring Bias Audit Law

Local Law 144 covers an automated employment decision tool used in the city when it substantially assists or replaces discretionary decision-making for hiring or promotion, including screening. According to the [NYC Department of Consumer and Worker Protection](https://www.nyc.gov/site/dca/about/automated-employment-decision-tools.page), before using a covered tool you need:

-   an independent bias audit completed within the past year
-   a published summary of the audit results
-   notice to candidates, given 10 business days before use through an allowed method such as the job posting, that says the tool is used and what qualifications or characteristics it assesses, with instructions for requesting a reasonable accommodation

Coverage depends on the job location, and DCWP's guidance extends it to some fully remote roles tied to a New York City office. A remote job attached to your Manhattan office can be covered even if the recruiter sits in New Jersey. The notice duty also turns on where the candidate lives, so log both.

The city's rules read "substantially assists" fairly narrowly, for example when the output is the only factor, outweighs other factors, or overrides a human judgment. That's why Team B's cutoff matters.

The biggest trap is outsourcing. A vendor can arrange the audit, but the employer still has to make sure a valid audit exists, the summary is published, and notice goes out. "Our vendor handles compliance" won't hold up.

## Colorado: The Rewritten ADMT Act

If your notes on Colorado come from coverage of the original AI act, they're stale. Senate Bill 26-189 repealed and reenacted those provisions as the Automated Decision-Making Technology Act, and the [Colorado Attorney General](https://coag.gov/ai/) says the revised provisions take effect January 1, 2027.

The law covers automated decision-making technology used to materially influence consequential decisions, and employment is one of them. Covered deployers have notice duties and must provide disclosures after an adverse outcome. Candidates get rights to access and correct inaccurate personal data, plus an opportunity for meaningful human review of an adverse decision based on covered technology.

Rulemaking is still shaping the mechanics. The practical read: if your tool can reject a Colorado candidate, you need a route for that person to reach a human who can actually change the result. Design that queue before the first request arrives.

## EU AI Act: High-Risk Hiring Systems

The EU AI Act treats certain AI systems used for recruitment, candidate selection, and employment decisions as high-risk. Classification depends on intended use, and there are exceptions, so not every HR feature lands in that bucket. Tools that filter applications or evaluate candidates are the ones to map first.

For covered high-risk systems, the Act sets duties for risk management, data governance, records and logging, information for users, accuracy and cybersecurity, and human oversight. Much of that sits with the provider. The employer, as deployer, has to use the system as instructed, assign competent people to oversee it, keep the logs it generates, and inform candidates and workers where the Act requires it.

Emotion recognition in the workplace is prohibited, with narrow medical and safety exceptions. That rules out tools claiming to read enthusiasm or honesty from a face or voice. Transcribing what a candidate said is different.

The Act is in force but applies in phases, and the timing for high-risk obligations has been politically and legally fluid. Don't assume every duty above already applies to your system. Map your systems now, then confirm the live application date and any transition rules for your system with counsel before deployment. The Act can also reach a US employer whose system output is used in the EU, such as hiring for a role in Madrid.

## How to Stay Compliant With AI Hiring Laws: The Five-File Packet

Small teams can't run a separate compliance project per jurisdiction. What works is one packet per tool, built before launch and kept current. If counsel, a regulator, or a candidate's lawyer asks how the tool is used, these five files answer it.

### 1\. Tool and use inventory

-   tool name, vendor, model or version, and which features are switched on
-   the ladder level for each workflow
-   job families, job locations, and candidate locations it touches
-   inputs it uses (resume, transcript, audio, video) and outputs it produces
-   which laws counsel says apply, and why

### 2\. Candidate notice, consent, and accommodation route

-   the exact notice text, where it appears, and when it goes out relative to the interview
-   consent capture where required, with a timestamp per candidate
-   a named contact and a tested process for accommodation or alternative-process requests
-   a written answer to "what happens if I don't consent?"

These [AI hiring disclosure templates](https://getkira.com/blog/ai-hiring-disclosure) give you notice and accommodation wording to adapt for each location.

### 3\. Human-review decision rule

Write down who reviews, what they see, and what they can change. Meaningful review means the reviewer sees the candidate's actual answers, has authority to override the tool, has enough time per candidate to form a view, and records a reason when they disagree.

A recruiter clicking "approve" on 300 auto-rejections in ten minutes is a rubber stamp, not human review. Track override rates. If they sit at zero for months, the review probably isn't real. This piece on the [AI interviewer and where humans still decide](https://getkira.com/blog/ai-interviewer) covers how to split the work.

### 4\. Vendor evidence file

-   the latest bias audit, with its date, scope, and data source
-   model documentation covering what the tool evaluates and what it ignores
-   retention and deletion terms, including backups
-   the subprocessor list and where data is stored
-   the vendor's commitment to tell you about model changes

Check that the audit covers the version and configuration you actually run. The same request goes to any [AI candidate screening platform](https://getkira.com/ai-candidate-screening), including Kira AI, which automates phone screens with one-way AI interviews.

### 5\. Decision and change log

-   launch date and who approved it
-   every change to the model, questions, score threshold, or job locations
-   what you re-checked after each change
-   deletion requests and completion dates
-   candidate review requests and their outcomes

Each change should trigger a re-review. Moving a threshold from 60 to 70 changes who gets rejected and may mean the existing audit no longer describes the tool you're running.

## Pre-Launch Checklist for AI Interview Tools

Run this before the first candidate goes through, and again after any meaningful change.

1.  Classify every workflow on the ladder. Reclassify the moment anyone adds a cutoff.
2.  List job locations and candidate residences, including remote roles tied to a city office.
3.  Get counsel to confirm which laws apply to each workflow.
4.  Confirm a current bias audit exists where required and its summary is published.
5.  Publish notice through an allowed channel with the required lead time.
6.  Capture consent before any AI video analysis for Illinois roles.
7.  Test the accommodation and alternative-process route with a dummy request.
8.  Write the human-review rule and name the reviewers.
9.  Test deletion end to end, including vendor copies and backups.
10.  Confirm the tool doesn't use ZIP codes or similar proxies and doesn't claim to read emotions.
11.  Open the decision and change log.
12.  Set a re-review date and list the changes that trigger an earlier one.

## Key Takeaways

-   AI interview laws turn on location, what the tool does, the data it uses, and how much it shapes who advances. The vendor's label doesn't decide coverage.
-   The same product can be advisory in one workflow and dispositive in another. Adding an auto-reject threshold raises exposure most.
-   Illinois has two layers: a video-specific consent law and a broader Human Rights Act rule covering AI in recruitment.
-   A vendor's bias audit or compliance claim doesn't transfer the employer's duties.
-   Meaningful human review needs access to answers, authority to override, and a record of overrides.
-   Build the five-file packet before launch, and re-review whenever the model, threshold, questions, or job locations change.

[↑ Back to top](https://getkira.com/blog/ai-interview-laws-for-employers#top)

On this page

1.  [Start With What the Tool Does, Not What It's Called](https://getkira.com/blog/ai-interview-laws-for-employers#start-with-what-the-tool-does-not-what-it-s-called)
    1.  [One tool, two workflows](https://getkira.com/blog/ai-interview-laws-for-employers#one-tool-two-workflows)
2.  [AI Interview Laws Compared: Illinois, NYC, Colorado, EU](https://getkira.com/blog/ai-interview-laws-for-employers#ai-interview-laws-compared-illinois-nyc-colorado-eu)
3.  [Illinois: Two Laws With Different Triggers](https://getkira.com/blog/ai-interview-laws-for-employers#illinois-two-laws-with-different-triggers)
    1.  [The Artificial Intelligence Video Interview Act](https://getkira.com/blog/ai-interview-laws-for-employers#the-artificial-intelligence-video-interview-act)
    2.  [The Human Rights Act amendment](https://getkira.com/blog/ai-interview-laws-for-employers#the-human-rights-act-amendment)
4.  [New York City: The Best-Known AI Hiring Bias Audit Law](https://getkira.com/blog/ai-interview-laws-for-employers#new-york-city-the-best-known-ai-hiring-bias-audit-law)
5.  [Colorado: The Rewritten ADMT Act](https://getkira.com/blog/ai-interview-laws-for-employers#colorado-the-rewritten-admt-act)
6.  [EU AI Act: High-Risk Hiring Systems](https://getkira.com/blog/ai-interview-laws-for-employers#eu-ai-act-high-risk-hiring-systems)
7.  [How to Stay Compliant With AI Hiring Laws: The Five-File Packet](https://getkira.com/blog/ai-interview-laws-for-employers#how-to-stay-compliant-with-ai-hiring-laws-the-five-file-packet)
    1.  [1\. Tool and use inventory](https://getkira.com/blog/ai-interview-laws-for-employers#1-tool-and-use-inventory)
    2.  [2\. Candidate notice, consent, and accommodation route](https://getkira.com/blog/ai-interview-laws-for-employers#2-candidate-notice-consent-and-accommodation-route)
    3.  [3\. Human-review decision rule](https://getkira.com/blog/ai-interview-laws-for-employers#3-human-review-decision-rule)
    4.  [4\. Vendor evidence file](https://getkira.com/blog/ai-interview-laws-for-employers#4-vendor-evidence-file)
    5.  [5\. Decision and change log](https://getkira.com/blog/ai-interview-laws-for-employers#5-decision-and-change-log)
8.  [Pre-Launch Checklist for AI Interview Tools](https://getkira.com/blog/ai-interview-laws-for-employers#pre-launch-checklist-for-ai-interview-tools)
9.  [Key Takeaways](https://getkira.com/blog/ai-interview-laws-for-employers#key-takeaways)

Every applicant gets an interview.

Kira-AI runs the first conversation and writes the scorecard. Free trial, no card.

[Start free](https://app.getkira.com/signup)

25interviews free

Every applicant screened by voice. No card needed.

[Start free](https://app.getkira.com/signup)

## Try Kira-AI on one vacancy

Free trial, no card, 25 interviews.

[Start free](https://app.getkira.com/signup)[See how it works →](https://getkira.com/how-it-works)

## Keep reading

-   ![Abstract AI interview workflow with automated paths and a human decision checkpoint](https://getkira.com/_next/image?url=https%3A%2F%2Fcdn.sanity.io%2Fimages%2F44aj5pkb%2Fproduction%2Fa4c9bffb12835968a9eab86c65cf9b448c51392d-2624x1632.webp%3Fw%3D960%26h%3D540%26fit%3Dcrop&w=3840&q=75)
    
    ### [AI Interviewer: What It Does and Where Humans Still Decide](https://getkira.com/blog/ai-interviewer)
    
    An AI interviewer can standardize early screening, but it shouldn't make the hiring decision. Use this automate, review, and never-delegate framework to set clear boundaries.
    
    August 1, 2026
    
-   ![Abstract layered disclosure cards with a consent check and human review path](https://getkira.com/_next/image?url=https%3A%2F%2Fcdn.sanity.io%2Fimages%2F44aj5pkb%2Fproduction%2F62244ac8d1605de57b27398996bfdaed0c87b45c-2624x1632.webp%3Fw%3D960%26h%3D540%26fit%3Dcrop&w=3840&q=75)
    
    ### [AI Hiring Disclosure: Notice, Consent, and Accommodation Templates](https://getkira.com/blog/ai-hiring-disclosure)
    
    Use this three-layer AI hiring disclosure workflow to explain screening tools clearly. Includes copy-paste notices, consent language, and an accommodation reply.
    
    August 11, 2026
    
-   ![Abstract scorecard, transcript, and AI evaluation nodes for automated interview evaluation](https://getkira.com/_next/image?url=https%3A%2F%2Fcdn.sanity.io%2Fimages%2F44aj5pkb%2Fproduction%2F81535f11f1498ec95a5b39fe8c57e50cdeba99a9-2624x1632.webp%3Fw%3D960%26h%3D540%26fit%3Dcrop&w=3840&q=75)
    
    ### [Automated Interview Evaluation: How AI Scores Candidates](https://getkira.com/blog/automated-interview-evaluation)
    
    Automated interview evaluation can speed up screening, but only when scores are tied to structured questions, job-related rubrics, and human review.
    
    July 15, 2026
    
-   ![Abstract illustration of AI-powered candidate screening with neural network nodes and filtering funnel shapes in blue-purple gradient](https://getkira.com/_next/image?url=https%3A%2F%2Fcdn.sanity.io%2Fimages%2F44aj5pkb%2Fproduction%2F2a60fcd48b771c200188487c6c6c42b339bcf4b6-2624x1632.webp%3Fw%3D960%26h%3D540%26fit%3Dcrop&w=3840&q=75)
    
    ### [AI Candidate Screening: The Complete Guide for Recruiters](https://getkira.com/blog/ai-candidate-screening-guide)
    
    Learn how AI candidate screening works, where it saves the most time, and how to implement it in your hiring workflow without losing the human touch.
    
    March 19, 2026