Do you have to tell candidates you use AI? The answer by tool and by place (2026)
On this page
- Do you have to tell candidates you use AI? It depends on what the tool does
- Which places require notice now, and which start in 2027?
- What about an AI note taker that only transcribes?
- Does it matter where the candidate is?
- What happens if you don't tell them?
- How do you tell candidates, in practice?
- Questions people ask
Not everywhere, but in enough places that most hiring teams should. As of November 2026 no federal law requires a private US employer to tell candidates it uses AI. New York City does, at least 10 business days before an automated employment decision tool screens a city resident; Illinois does for AI used in hiring decisions, and requires notice and consent before AI analyzes recorded video interviews; Maryland requires a signed waiver for facial recognition in interviews; and California, Colorado and Connecticut add notice duties in 2027. In the EU and UK, data protection law already requires you to explain how candidate data is processed, and wherever an AI note taker records the call, recording-consent law applies whatever the AI does.
This page answers the question by what your tool does, then by place and date. For the full statute-by-statute map, see AI hiring laws by state; for wording you can copy, see the AI disclosure templates.
This is not legal advice. Each rule below was checked against the statute, regulation or regulator page linked next to it, as of November 2026. Several of these laws have rules still being written, and whether a particular tool is covered depends on how it works and how your team uses its output. Confirm with counsel for the places you hire.
Do you have to tell candidates you use AI? It depends on what the tool does
The AI notice laws are written around what a tool's output does to the decision, not around the word "AI". A tool that writes down what was said is in a different position from one that scores, ranks or rejects people. Find the row that matches your tool.
| What the tool does | Do AI notice laws reach it? | What applies anyway |
|---|---|---|
| Transcribes the interview or takes notes | Mostly not. New York City's rule says its target is not "the output from analytical tools that translate or transcribe existing text, e.g., ... transcribe a video or audio interview" (DCWP final rule). Colorado's new law excludes tools that summarize information for later human review (SB 26-189). Illinois is unsettled: see below. | Recording consent in all-party consent states; privacy notices (the CCPA for California applicants, the GDPR and UK GDPR) |
| Summarizes the interview or drafts a scorecard a person reviews and can change | Usually treated like notes while a person really makes the decision. It moves toward coverage when a draft score is what actually decides: Connecticut's test is a "substantial factor", Colorado's is output that "materially" influences the decision. | As above; keep the human review genuine and documented |
| Screens, scores, ranks or recommends candidates | Yes in New York City (bias audit and notice), Illinois (notice), and from 2027 in California, Colorado and Connecticut; in the EU from 2 December 2027 as a high-risk system | Title VII disparate-impact rules, whoever built the tool |
| Analyzes applicant-recorded video interviews | Yes for Illinois-based positions: notice, an explanation of how the AI works and the characteristics it evaluates, and consent, all before the interview (820 ILCS 42/5) | Deletion within 30 days of the applicant's request (820 ILCS 42/15) |
| Uses facial recognition during an interview | Yes in Maryland: a signed waiver stating the applicant's name, the interview date, their consent and whether they read the waiver (Md. Lab. & Empl. § 3-717) | Biometric privacy laws in other states |
| Talks to candidates (a chatbot or AI interviewer) | Connecticut, for deployments from October 1, 2027, where the tool also feeds the hiring decision: tell each applicant they are interacting with the technology, unless that is obvious. In the EU, Article 50(1) of the AI Act has applied since 2 August 2026 | Every scoring rule above, if it also scores answers |
Which places require notice now, and which start in 2027?
The dates matter because several duties begin within the next 13 months. This is the timeline as it stands in November 2026.
| In force | Where | What you must tell candidates | Source |
|---|---|---|---|
| Since 2020 | Illinois (video interviews) | Notice, explanation and consent before AI analyzes a recorded video interview | 820 ILCS 42/5 |
| Since 2020 | Maryland | Signed waiver before facial recognition creates a facial template in an interview | § 3-717 |
| Since July 5, 2023 | New York City | That an automated employment decision tool is used, the qualifications and characteristics it assesses, and how to request an accommodation, 10 business days before use | DCWP FAQ |
| Since January 1, 2026 | Illinois (all AI in employment decisions) | Notice that AI is used for recruitment, hiring and other listed decisions; the rules on timing and means are not final | 775 ILCS 5/2-102(L) |
| Since February 5, 2026 | United Kingdom | Safeguards for solely automated significant decisions, including information about the decisions | Data (Use and Access) Act 2025, s. 80 |
| January 1, 2027 | California | A Pre-use Notice before using automated decisionmaking technology for a hiring decision, plus opt-out and access rights | CCPA regulations §§ 7200, 7220 |
| January 1, 2027 | Colorado | Notice before a covered tool materially influences the decision; a plain-language disclosure within 30 days after an adverse outcome | SB 26-189 |
| October 1, 2027 | Connecticut | Disclosure of interaction, and a written notice before the decision with six items, including the tool's trade name and the data it analyzes | P.A. 26-15, §§ 9-10 |
| 2 December 2027 | European Union | Deployers of high-risk systems, which include recruitment tools, must inform the people subject to their decisions (AI Act Art. 26(11)) | Regulation (EU) 2026/1744 |
Colorado's date comes from the legislature's fiscal note, which says the act's other sections take effect on January 1, 2027 and apply to decisions made on or after that date; the attorney general must also adopt rules. The detail for each state is in the guides to Local Law 144, Illinois HB 3773, Colorado's law and California's ADMT rules.
Two places people ask about are not on the list. Texas's TRAIGA (HB 149) requires AI disclosure from government agencies and health care providers, not private employers. Utah's generative AI disclosure duty applies on request in a consumer transaction, and whether a job interview is one is unsettled.
What about an AI note taker that only transcribes?
This is the question most recruiters are really asking, and the honest answer has three parts.
- The AI notice laws mostly do not target it. New York City's rule excludes transcription output, Colorado excludes tools that summarize for human review, and California's rules cover technology that replaces or substantially replaces human decisionmaking. A transcript that a person reads before deciding does neither.
- Illinois is the exception to watch. Its notice duty reaches an employer "using artificial intelligence" for recruitment and hiring, the definition expressly includes generative AI, and the Department of Human Rights postponed its rulemaking on how to give notice on June 2, 2026, according to the Illinois Human Rights Commission's June 2026 presentation. We found no final rules as of November 2026, so telling Illinois candidates is the lower-risk reading.
- Recording law applies whatever the AI does. A note taker captures the conversation, and states such as California require the consent of all parties to record a confidential communication (Penal Code § 632). The full list is in two-party consent states. Privacy law adds a notice duty too: GDPR Article 13 requires you to tell candidates the purposes of processing when you collect their data (Regulation (EU) 2016/679).
In practice the answer is the same as for any recording: say it at the start of the call and give the candidate a real choice. An example line: "Before we start, I use an AI note taker that transcribes our conversation so I can listen instead of typing. It doesn't score you; the hiring team does that. Are you comfortable with that?" More versions, including what to do when someone says no, are in the interview recording consent script, and options for candidates who dislike meeting bots are in AI note takers when candidates object to bots.
Does it matter where the candidate is?
Yes, and each law draws the line differently, so a remote hiring process can trigger several at once.
- New York City requires notice to candidates who are city residents, and the law applies when the job is tied to a city office or the employment agency is in the city (DCWP FAQ).
- Illinois' video act applies to positions based in Illinois, wherever the candidate is.
- Connecticut covers an employee or applicant "in the state", and the deployer must be doing business there.
- California's CCPA rules protect California residents, including job applicants.
- Recording consent can depend on both people's states; when you and the candidate are in different ones, plan for the stricter rule.
What happens if you don't tell them?
Enforcement differs by law. In New York City, the Department of Consumer and Worker Protection takes complaints through 311 and its website, penalties run up to $500 for a first violation and $500 to $1,500 for each later one, and each missing notice to a candidate is a separate violation (Local Law 144, § 20-872). In Illinois, failing to give notice is itself a civil rights violation under the Human Rights Act, separate from any discrimination claim. Connecticut's and Colorado's duties are enforced only by their attorneys general, as unfair or deceptive trade practices, and neither law creates a new private lawsuit. Colorado's attorney general must offer a 60-day cure period before suing until 2030; Connecticut's may offer one for violations through December 31, 2027. Recording without the consent a state requires is a separate exposure, under wiretap statutes that carry their own penalties.
The practical cost comes first, though. A bot in the participant list or a transcript in a rejection dispute tells the candidate anyway, and finding out after the fact reads worse than a one-line notice at the start.
How do you tell candidates, in practice?
- List each tool and what its output does. Transcribes, summarizes, scores, ranks, analyzes video, or talks to candidates. The row in the first table decides which rules apply.
- Map your hiring locations against the timeline above, including where remote candidates live.
- Put a short statement on your careers page and a line in job postings where a scoring tool is used. In New York City, the careers-page notice starts the 10-business-day clock.
- Repeat it in the interview confirmation email and say it again at the start of the call, with a pause or an alternative on offer.
- Log that notice was given with the candidate record, and review the wording before the 2027 dates.
Some note takers build the spoken part in. Interview Signal, which we make, has a read-aloud consent line, notice text by jurisdiction and a one-click pause; like any tool's wording, it is not legal advice. Whatever you use, the templates in AI disclosure to candidates cover the careers page, posting, email and spoken versions.
Questions people ask
Do I have to tell candidates I'm using an AI note taker in the interview?
Usually not because of an AI law, but almost always because of recording and privacy law. The AI notice rules in New York City, Colorado and California are written around tools that score, rank or decide, and New York City's rule says transcription output is not what it targets. A note taker still records the conversation, so all-party consent states such as California require everyone's consent, and privacy notices under the CCPA or the GDPR must describe the processing. Illinois is the open question, because its notice duty covers AI used with respect to recruitment and hiring and has no final rules, so telling candidates there is the safer course.
Is there a federal law that requires employers to disclose AI use to job candidates?
No, as of November 2026 no federal statute or regulation requires a private employer to tell job candidates that it uses AI. Federal anti-discrimination law, including Title VII and the ADA, still applies to any selection procedure an AI tool produces, and the ADA's accommodation duty is easier to meet when candidates know a tool is being used. The notice duties that exist are state, city, EU and UK rules.
Do I have to tell a candidate if I used ChatGPT to screen resumes?
It depends on where you hire and how much the output decides. Using a chatbot to rank or filter resumes can make it an automated employment decision tool in New York City, which needs a bias audit and notice 10 business days before use, and it is use of AI in hiring under Illinois' Human Rights Act, which requires notice. California, Colorado and Connecticut add notice duties for tools that make or materially influence hiring decisions from 2027. Using a chatbot to draft a job description is not something these laws target.
How early do I have to tell candidates in New York City?
At least 10 business days before you use an automated employment decision tool on a candidate who lives in the city. DCWP's FAQ says the notice can go in a job posting or by mail or email, or, for applicants, on the employment section of your website, and an employer can start using the tool 10 business days after the website notice is posted regardless of when a particular job was posted.
Can a candidate refuse to be evaluated by AI?
In some places. Illinois forbids AI analysis of a recorded video interview without the applicant's consent, New York City's notice must allow a candidate to request an alternative selection process or accommodation, and California's rules make an opt-out from automated decisionmaking the default from January 1, 2027, with listed exceptions such as offering an appeal to a human reviewer. Elsewhere there is no general right to refuse, but offering a human alternative is good practice.