Is Video Interview Screening Legal? NYC Local Law 144, the EU AI Act, Illinois, and What Actually Triggers Them

Updated 2026-10-16: added Colorado's SB 26-189 and Connecticut's CART Act (SB 5), both taking effect in 2027 — see "Colorado and Connecticut join the list" below. Updated 2026-10-15: added a new section on Illinois's amendment to the Human Rights Act (Public Act 103-0804 / HB 3773), effective January 1, 2026 — broader than the state's older, video-specific AI Video Interview Act — see "What Illinois's 2026 amendment adds" below. Updated 2026-09-11: the EU AI Act deferral below is now final, not pending — see "What the EU AI Act actually regulates." Also updated 2026-09-03 with two live 2026 lawsuits against algorithmic screening vendors — see "Real litigation, not a hypothetical" below.

Two hiring-tech laws get cited constantly and understood rarely: NYC's Local Law 144 and the EU AI Act. Both regulate automated hiring decisions — not video interviews as a format. The distinction matters a lot if you're evaluating a screening tool in 2026, because it decides whether you need an independent bias audit and a compliance program, or nothing at all.

This isn't legal advice — it's a plain-English map of what each law actually covers, so you know which questions to ask a vendor (or your own counsel) before you buy.

What NYC Local Law 144 actually regulates

Local Law 144 covers AEDTs — Automated Employment Decision Tools: software that uses machine learning, statistical modeling, or AI to substantially assist or replace a hiring or promotion decision. Think resume rankers, scored assessments, and AI systems that score or rank interview answers.

If a tool is an AEDT, NYC employers using it on candidates working in the city must:

• Get an independent bias audit done annually, by a third party — not the employer, not the vendor — assessing the tool for bias across sex and race/ethnicity categories, with a summary published publicly.
• Notify candidates at least 10 business days before using the tool, stating that an AEDT is in use, what it assesses, and (on request) its data source and retention policy.
• Face penalties starting at $500 per violation, escalating to $1,500/day for ongoing non-compliance.

The trigger is the word "substantially" — a tool that collects information for a human to review isn't automating the decision; a tool that outputs a score or rank the employer relies on is.

What the EU AI Act actually regulates

The EU AI Act lists recruitment and employment systems in Annex III as high-risk: CV filtering, candidate ranking, performance evaluation, and termination-related decisions. High-risk systems require risk management, bias testing, logging, and human oversight. Deployers (the businesses using the tool), not just the vendors that built it, share responsibility for meeting those obligations.

The compliance timeline has moved, and — as of this update — that move is now final: the original August 2, 2026 deadline for high-risk employment AI was pushed to December 2, 2027, via the EU's "Digital Omnibus on AI." The Council of the EU gave its final green light on June 29, 2026, following the European Parliament's formal approval, and the Omnibus entered into force on July 27, 2026 after publication in the Official Journal of the EU on July 24, 2026. So if you read an older source citing August 2026 as a hard deadline, it's out of date — the real deadline for high-risk employment AI obligations is now December 2, 2027.

Same underlying logic as Local Law 144: the law targets systems that make or meaningfully shape the hiring decision, not any tool that touches a candidate's data.

What Illinois's 2026 amendment adds (and why it's broader than the other two)

Illinois already had a narrow, video-specific law — the Artificial Intelligence Video Interview Act (820 ILCS 42), in force since 2020, which only applies when AI actually analyzes a candidate's recorded video (facial expression, word choice, tone). That's the same "AI analyzes the response" trigger as the EU AI Act and Local Law 144, and manually-reviewed video screening falls outside it for the same reason.

What's new, and materially broader: Public Act 103-0804 (HB 3773), effective January 1, 2026, amends the Illinois Human Rights Act itself — separate from, and on top of, the video-specific law above. It requires employers to give notice whenever AI is used in recruitment, hiring, promotion, discipline, discharge, or other employment decisions, and defines "AI" broadly as any machine-based system generating "predictions, content, recommendations, or decisions" — not just scoring or ranking software. It also prohibits AI that has a discriminatory effect on a protected class, including using a ZIP code as a proxy for a protected characteristic. The Illinois Department of Human Rights enforces it and is still adopting implementing rules.

The practical difference: Local Law 144 and the EU AI Act both turn on whether software substantially assists or replaces the hiring decision. Illinois's new amendment doesn't use that same narrowing language — it's written to cover AI use in the recruitment and hiring process more broadly, which is a lower bar to trigger notice obligations than the "does it score or rank candidates" test the rest of this article walks through. If you're hiring Illinois-based candidates, that's a real reason to run your own screening stack past counsel even if a vendor (Flash Interview included) doesn't score or rank anyone — "no AI decision-making" is the test for Local Law 144 and the EU AI Act; it may not be the full test under Illinois's broader language.

Colorado and Connecticut join the list (both 2027, both narrower than Illinois)

Two more states have AI-employment notice laws coming, and — unlike Illinois's amendment above — both use the same narrow trigger as NYC Local Law 144 and the EU AI Act, not Illinois's broader "any AI use" standard.

Colorado originally passed a broad AI Act (SB 24-205) that would have imposed proactive risk-management duties on any "consequential decision," including employment. Before it ever took effect, Colorado's legislature repealed those provisions and replaced them with a narrower bill, SB 26-189, signed May 14, 2026. The replacement pushes the effective date to January 1, 2027 and reorients the law around consumer transparency and rights tied to "consequential automated decision-making" — the same substantially-assists-the-decision framing as Local Law 144, not Illinois's lower bar.

Connecticut signed its own law — SB 5, the Artificial Intelligence Responsibility and Transparency ("CART") Act — on May 27, 2026. It defines "automated employment-related decision technology" (AEDT) as a system whose output (a score, ranking, classification, or recommendation) is a "substantial factor" in, or materially influences, an employment decision — explicitly excluding ordinary productivity software or tools used only incidentally. The point-of-interaction and pre-decision notice requirements both take effect October 1, 2027, and the law separately states that using an AEDT is not a defense to a discrimination claim.

The practical read for a video-screening buyer: Colorado and Connecticut both ask the same question Local Law 144 and the EU AI Act already ask — does the software's output substantially drive the decision, or does a person? Illinois's HB 3773 (above) is still the outlier with the lower, any-AI-use bar. A tool like Flash Interview, which records video and hands it to a human to watch and judge with no score, rank, or recommendation attached, falls outside the Colorado and Connecticut triggers for the same structural reason it falls outside Local Law 144 and the EU AI Act — but as with Illinois, confirm your own hiring stack with counsel rather than relying on any vendor's characterization, including ours.

The question that actually matters: who makes the decision?

Both laws are aimed at the same thing from different angles — algorithmic decision-making that replaces human judgment in who gets hired. That means the honest question to ask about any screening tool isn't "does it use AI anywhere in the product," it's:

Does a person, or the software, decide who moves forward?

• A tool that scores, ranks, or auto-shortlists candidates from their video or resume is doing the thing both laws regulate — you should expect (and demand) a bias audit trail and a notice process from that vendor.
• A tool that records candidate video and hands it to your team to watch and judge isn't making an automated decision — there's no algorithmic output for a regulator to audit, because there isn't one.

Flash Interview is the second kind: we collect the video and put it in front of your hiring team. We don't score, rank, or auto-shortlist anyone. Your team watches the answer and decides, the same as if the candidate had sent you a video file directly. That's a real, structural difference from AI-scoring platforms — not a compliance claim we're making, just what the product does.

Real litigation, not a hypothetical

The scenario both laws are written for — software that scores, ranks, or filters candidates before a person sees them — isn't theoretical anymore. Two cases moved forward in early 2026:

• Mobley v. Workday reached a milestone in February 2026 when a federal court authorized notice to a nationwide class of applicants who allege Workday's AI-driven screening software filtered them out based on age, race, and disability. The court treated the software vendor itself as an "agent" of the hiring employers using it — meaning an employer can't point at the vendor and walk away from liability for what the tool did.
• Kistler v. Eightfold AI (filed January 2026) is a proposed class action alleging Eightfold's platform scored job applicants on a 0–5 scale using data pulled from third-party sources — without disclosing the score to the candidate or following the Fair Credit Reporting Act's disclosure requirements for that kind of report. Notably, the claim isn't that the score was biased; it's that the scoring existed at all, undisclosed.

Two different theories — discrimination in one, undisclosed consumer-report scoring in the other — but the same underlying fact pattern both times: a piece of software silently decided who advanced before a human looked at the file. That's exactly the trigger both Local Law 144 and the EU AI Act (and now, apparently, the EEOC and the plaintiffs' bar) are aimed at.

Flash Interview doesn't do that. There's no score, no rank, no algorithmic shortlist for a court to examine — a person on your hiring team watches the recorded answer and decides, the same as watching any video a candidate sent you directly. That's not a compliance claim, it's just what the product does, and it's the same structural difference that keeps you out of Local Law 144's AEDT definition (see above).

What to ask any video-screening vendor

• "Does your product output a score, rank, or recommendation on candidates?" If yes, ask about their bias-audit cadence and candidate-notice process.
• "Who makes the shortlist decision — your algorithm or my team?"
• "If I'm hiring in NYC or the EU, what does my Local Law 144 / AI Act obligation look like with your tool specifically?"

None of this is a substitute for your own counsel, especially if you're hiring at volume in a regulated jurisdiction or in a regulated industry (financial services, healthcare). Laws in this space are moving fast — the EU AI Act deadline slip above (now finalized, after months as a proposal) is a good example of how much these dates can move before they settle. Confirm current status before you rely on it.

Frequently asked questions

Does one-way video interview screening count as an AEDT under NYC Local Law 144?

Only if the tool scores, ranks, or otherwise substantially assists the hiring decision algorithmically. A tool that just records and delivers video for a human to review isn't automating the decision, so it falls outside the AEDT definition — but confirm this with your own counsel for your specific setup, since "substantially assist" is a judgment call the law doesn't fully spell out.

Is manually-reviewed video screening covered by the EU AI Act's high-risk rules?

The Annex III high-risk category targets systems that filter, rank, or evaluate candidates algorithmically. A video collection tool with no scoring or ranking output doesn't fit that description, but the deployer (you) still carries responsibility for classifying your own hiring stack correctly — don't rely on a vendor's marketing claim alone.

When does the EU AI Act's high-risk employment AI requirement actually take effect?

December 2, 2027. The original August 2, 2026 deadline was deferred via the EU's Digital Omnibus on AI, which received final Council approval on June 29, 2026 and entered into force on July 27, 2026 — the deferral is now settled law, not a pending proposal.

What's the penalty for using an unaudited AEDT under Local Law 144?

Penalties start at $500 per violation and escalate to $1,500 per day for continued non-compliance — a strong reason to know definitively whether your screening tool is an AEDT before you use it on NYC-based candidates.

Are there real lawsuits over AI hiring screening, or is this still theoretical?

It's real. In February 2026, a federal court authorized class notice in Mobley v. Workday, where applicants allege Workday's screening AI filtered them out by age, race, and disability. In January 2026, a separate proposed class action (Kistler v. Eightfold AI) alleged an AI hiring platform scored candidates without disclosing it, raising Fair Credit Reporting Act claims. Neither case involves manually-reviewed video screening — both target software that scores or ranks candidates algorithmically before a human sees them.

Does Illinois have its own AI hiring law, separate from Local Law 144 and the EU AI Act?

Yes, two of them. The older Artificial Intelligence Video Interview Act (820 ILCS 42, in force since 2020) only applies when AI analyzes a candidate's recorded video — the same narrow trigger as Local Law 144 and the EU AI Act. The newer one, Public Act 103-0804 (HB 3773), effective January 1, 2026, amends the Illinois Human Rights Act and is broader: it requires notice whenever AI is used anywhere in recruitment or hiring, not only when it scores or ranks candidates. If you're hiring Illinois-based candidates, confirm with counsel whether your full hiring stack — not just the video-screening step — triggers notice under the newer, broader law.

What do Colorado and Connecticut require, and when?

Colorado's SB 26-189 (signed May 14, 2026, replacing an earlier, broader bill that never took effect) takes effect January 1, 2027, and covers "consequential automated decision-making" — the same substantially-assists-the-decision trigger as Local Law 144. Connecticut's SB 5, the CART Act (signed May 27, 2026), requires notice when an "automated employment-related decision technology" output is a substantial factor in, or materially influences, an employment decision; both of its employee/applicant notice requirements take effect October 1, 2027. Neither uses Illinois's broader "any AI use" standard — both ask whether the software's output drives the decision, which is the same question this whole article walks through.

---

Sources

• Mobley v. Workday class notice authorized, February 2026 — Fisher Phillips
• Kistler et al. v. Eightfold AI Inc., FCRA class action filed January 2026 — Akin Gump
• AI Hiring Under Fire: What the Eightfold Lawsuit Means for Every Employer — Jones Walker LLP
• EU AI Act Omnibus Agreement — Postponed High-Risk Deadlines and Other Key Changes — Gibson Dunn
• HR Tools and Artificial Intelligence: Europe Delays High-Risk Obligations to December 2027 — ActuIA
• Illinois Enacts Artificial Intelligence Law Focused on Employment Practices — Duane Morris LLP
• Colorado Delays Comprehensive AI Law With Further Changes Anticipated — Greenberg Traurig
• Connecticut Takes Aim at AI in Employment Decisions With SB 5 — Alston & Bird
• Connecticut's New AI Law Mandates Disclosure of Automated Employment-Related Decision Technology — Paul Hastings