How to Write an AI Disclosure Notice for Candidates

Multiple jurisdictions now require employers to disclose when AI is used in hiring decisions. A well-crafted disclosure notice protects your organization, builds candidate trust, and satisfies legal requirements. This guide provides templates you can customize for your needs.

Key Principles for Effective Disclosures

• Clear language: Avoid technical jargon; write for a general audience
• Specific: Identify the actual tools and purposes, not vague generalities
• Complete: Cover all required elements for applicable jurisdictions
• Timely: Provide notice before AI is used on the candidate
• Accessible: Make notices easy to find and save
• Actionable: Include contact information and explain candidate rights

What Must Be Disclosed

Requirements vary by jurisdiction, but common elements include:

Universal Template

This template covers the most common requirements and can be customized for your organization:

NYC Local Law 144 Template

This template specifically addresses NYC's AEDT requirements:

Colorado AI Act Template

Colorado requires more emphasis on consumer rights and appeal processes:

Where to Place Notices

Ensure candidates receive notices through multiple touchpoints:

• Job postings: Include notice or link in every job posting
• Career site: Prominent disclosure on careers/jobs page
• Application confirmation: Email sent after application submission
• Before assessments: Display notice before AI-powered tests
• Privacy policy: Include in employment-related privacy disclosures

Customization Checklist

Before using these templates, customize for your organization:

• ☐ Insert your company name
• ☐ List specific AI tools by name
• ☐ Describe actual purposes (don't use generic language)
• ☐ Specify actual data inputs for your tools
• ☐ Describe actual outputs (scores, rankings, etc.)
• ☐ Add your contact information
• ☐ Link to bias audit summary (NYC)
• ☐ Confirm opt-out process is operational
• ☐ Review with legal counsel

Frequently Asked Questions

Can I use one disclosure template for all jurisdictions, or do I need separate notices?

You can create one comprehensive disclosure that meets the requirements of all applicable jurisdictions (the "highest common denominator" approach). Include all elements required by the strictest laws (Colorado's opt-out rights, NYC's 10-day notice, Illinois consent, California ADMT details) and deliver to all candidates nationally. This simplifies operations. However, some employers prefer jurisdiction-specific notices to avoid over-disclosure in states with minimal requirements. Trade-off: simplicity vs. precision. For most employers, one comprehensive notice is easier to maintain and reduces the risk of sending the wrong template to the wrong candidate.

How should we deliver AI disclosure notices—email, job posting, application page, or all three?

Best practice: multi-channel delivery. Include basic disclosure in job postings ("This employer uses AI in hiring. Learn more at [link]"), send dedicated notice via email when candidate applies, and repeat before AI assessment stages. This ensures notice regardless of how candidates found your job (direct apply, job board, referral). For NYC's 10-day advance notice requirement, the job posting or application confirmation email provides the earliest timestamp. For Illinois consent requirements, in-app disclosure with checkbox before AI assessment works best. Document delivery across all channels—if a candidate claims they never received notice, you want multiple proof points.

Do we need separate disclosures for each AI tool, or can one notice cover multiple tools?

One notice can cover multiple AI tools if it describes each one specifically. Example: "We use the following AI tools in our hiring process: (1) Workday Recruiting for resume screening and candidate matching, (2) HireVue for video interview analysis, and (3) Codility for technical skills assessment." However, don't use overly broad language like "we may use various AI tools"—this lacks the specificity required by laws like Colorado and Illinois. If you add new AI tools mid-hiring cycle, send supplemental disclosure before using them. See our Compliance Program Guide for managing multi-tool disclosure workflows.

What's the difference between disclosure and consent? Do we need both?

Disclosure means informing candidates that AI will be used and how it works. Consentmeans obtaining their affirmative permission to use AI on their application. Illinois explicitly requires both disclosure and consent. NYC and Colorado require disclosure but don't mandate affirmative consent (though they require opt-out options, which is related). California CCPA ADMT requires disclosure and opt-out ability. Best practice: obtain explicit consent everywhere via checkbox or signature, even where not strictly required—this demonstrates good faith and reduces argument that candidates didn't understand their rights. Document consent with timestamps and retain for statute of limitations period (typically 2-4 years).

Can candidates opt out after we've already used AI on their application?

Yes, and you should honor retroactive opt-outs. If a candidate learns you used AI (perhaps they didn't see the disclosure, or didn't understand it) and requests human-only review, conduct fresh evaluation without relying on AI outputs. This is both legally safer and better candidate experience. Colorado's AI Act requires employers to "provide a mechanism" for candidates to appeal AI-influenced decisions—retroactive opt-out and fresh review satisfies this. Document the alternative evaluation process and ensure the candidate isn't penalized for requesting it.

How often should we update our disclosure templates?

Review and update quarterly or whenever: (1) you add new AI tools, (2) AI vendors update their algorithms (material changes), (3) applicable laws change, or (4) you expand hiring to new jurisdictions with different requirements. Many employers set calendar reminders for quarterly compliance reviews. Use version control—date each template version and track which candidates received which version. This is critical if you're audited or sued: you need to prove what candidates were told at the time of their application. Store historical template versions for at least 4 years (NYC bias audit retention requirement).

What if our AI vendor won't provide enough information for us to create a detailed disclosure?

That's a major red flag. If your vendor can't or won't explain how their AI works, what it analyzes, and how outputs are generated, you can't meet disclosure requirements in states like Colorado and California. More fundamentally, you can't validate the tool for bias or ensure it's not discriminatory—which creates enormous legal exposure under federal and state anti-discrimination laws. Options: (1) Demand transparency from vendor with threat to switch, (2) Conduct independent analysis/reverse engineering (expensive and difficult), or (3) Switch to a vendor that provides adequate documentation. See our Vendor Assessment Guide for compliance support requirements to include in AI vendor contracts. The EEOC has made clear: "I didn't know how my vendor's AI worked" is not a defense for discrimination.

Related Resources

• Do I Need to Disclose AI? Decision Tree
• What Counts as AI in Hiring?
• NYC Local Law 144 Guide
• Illinois HB 3773 Guide
• Free Compliance Scorecard