AI Hiring Bias Audit: What Candidate Notice Rules Mean

On 1 October 2026, Connecticut’s law on automated employment-related decision technology took effect, requiring employers to give written notice when tools inform employment decisions. It arrives in a year when AI hiring rules have become a patchwork: some places demand an independent AI hiring bias audit, others ask only for notice and explanation, and many proposals have stalled. For any company that uses AI in interviews, the practical question is simple. Which obligations apply, and what evidence will you need if a candidate or regulator asks questions?

This guide walks through the main rules that exist today, what they mean for interview platforms, and how a hiring team can prepare without slowing down honest candidates.

Where the rules stand in late 2026

The clearest rule for bias testing is in New York City. Local Law 144 has been enforced since 5 July 2023 and applies to employers and employment agencies that use an automated employment decision tool. According to the NYC Department of Consumer and Worker Protection, a tool may not be used unless it has been subject to a bias audit within one year before use, the audit information must be publicly available, and candidates must receive notice before the tool is used.

Elsewhere the emphasis has shifted. A MultiState review published on 21 September 2026 notes that several bills requiring bias or impact assessments died in 2026, including New Jersey’s A 2726, which would have required annual independent bias audits. The enacted laws it describes focus instead on telling applicants when a tool is used, explaining adverse outcomes, and allowing data to be corrected. Colorado and Connecticut are both in this group.

HR Dive’s running list of state and local rules, last updated on 16 June 2026, gives the dates that matter most for employers:

• Connecticut: the automated employment-related decision technology law took effect on 1 October 2026 and requires written notice covering the tool’s purpose, data categories and data sources. Disclosure duties for covered tools begin for uses on or after 1 October 2027.
• Colorado: SB 26-189 takes effect on 1 January 2027. Employers must notify people when a tool is used for consequential decisions, describe its role within 30 days of an adverse outcome, allow a request for human review, and keep records for at least three years.
• California: employers covered by the CCPA regulations face pre-use notice and opt-out or appeal duties from 1 January 2027. Civil Rights Council rules, in effect since 1 October 2025, already bar automated systems that discriminate on a protected basis and require applicant decision data to be kept for at least four years.
• Illinois: the AI Video Interview Act, in force since 1 January 2020, requires notice and consent before AI analyses a video interview, and lets candidates ask for their video to be deleted within 30 days.

The HR Dive list is a secondary summary, so employers should confirm each date against the official statute or regulation before relying on it.

Why the audit question is still the hard one

A bias audit asks whether a selection tool produces different outcomes for different groups. NYC’s rules require calculations such as selection and scoring rates and impact ratios, and they leave open how far the audit must reach beyond the tool itself. An audit is a statistical exercise run on outcome data. It needs enough candidates, clean records of who was scored and what decision followed, and an independent reviewer.

Litigation shows why this matters. In Mobley v. Workday, applicants allege that Workday’s AI screening tools screen out older, Black and disabled applicants. According to an Interviewer Signal review of AI hiring lawsuits, the court in July 2024 allowed disparate impact claims to proceed and treated a third-party vendor as a possible agent of employers. Plaintiffs moved to certify a class on 14 September 2026, and Workday’s opposition is due on 10 November 2026, with a hearing set for 9 March 2027. No court has found Workday liable. The case is still a warning that vendors and employers can both face questions about how a tool affects candidates.

Another pending case, Kistler v. Eightfold AI, asks whether AI candidate scoring should be treated as a consumer report. As described in the same review, the motion to dismiss was pending with no ruling reported in the material available at the time of writing.

What a defensible AI interview process looks like

Whatever rules apply in your state, several practices make an AI-assisted interview easier to defend:

• Tell candidates early. Explain that AI is used, what it does, and that a person makes the decision.
• Give the same structure to everyone. Use the same questions, the same scorecard and the same round logic for each candidate at a given level.
• Keep the evidence together. Store the transcript, the recording where consented, the scorecard and the final decision in one record.
• Review flags before acting on them. Any integrity signal, such as a second face or a tab switch, is a prompt for human review, not a verdict.
• Measure outcomes by group. Where the law requires an audit, or the employer wants one, track selection rates and review them with someone independent of the hiring team.

How InstaDataAssess supports fair and reviewable hiring

InstaDataAssess is an AI and expert interview platform built on one principle: AI does the volume, and your people make the calls. AI interviewers run voice-first screening rounds and coding tasks, with answers graded on both the solution and the reasoning. Shortlisted candidates move to live expert video rounds, where scorecards are drafted for HR to approve.

On the record side, every round produces an evidence-rich report. It includes the per-skill scores, the question-by-question transcript, the recording and an integrity timeline. The Decision Cockpit combines AI and expert scores into a single advance, hold or hire recommendation with the evidence shown beside it. Nothing is auto-rejected. Invites are gated by HR, each level allows one decision, and a complete audit log records what was done.

It is important to be clear about limits. InstaDataAssess helps you keep the records that an audit or a candidate query will need. It does not replace an independent bias audit, and it does not decide whether your use of AI meets a particular state’s rules. That judgement belongs to your legal and compliance team, working from the official text.

For more practical guides on hiring integrity and compliance, visit the InstaDataHelp blog.

Frequently Asked Questions

What is an AI hiring bias audit?

It is a statistical review of whether an automated hiring tool produces different selection or scoring outcomes for groups defined by characteristics such as sex, race or ethnicity. New York City’s Local Law 144 requires such an audit within one year before a covered tool is used.

Do all US states require AI hiring bias audits?

No. Based on the MultiState review of September 2026, many bills that would have required audits died, and the enacted laws focus more on notice, explanations and human review. Employers still need to check the rules in each state and city where they hire.

What notice must a candidate receive before an AI interview?

It depends on the jurisdiction. NYC requires notice at least 10 business days before a covered tool is used. Connecticut requires written notice of the tool’s purpose, data categories and data sources. Illinois requires notice and consent before AI analyses a video interview.

Can AI reject a candidate on its own under these rules?

The rules described above generally require an employer to remain responsible for the decision, and Colorado’s law gives applicants a route to request human review of an adverse outcome. In InstaDataAssess, recommendations are reviewed by people and nothing is auto-rejected.

Sources

• NYC Department of Consumer and Worker Protection: Automated Employment Decision Tools
• HR Dive: A running list of states and localities that regulate AI in hiring (updated 16 June 2026)
• MultiState: State AI employment laws address monitoring, hiring and bias concerns (21 September 2026)
• Interviewer Signal: AI hiring lawsuits in 2026, including Mobley v. Workday

Build a hiring process you can explain to a candidate and defend to a regulator. Explore how InstaDataAssess combines AI screening, expert panels and a human-led decision, or write to info@instadatahelp.com to arrange a walkthrough.