The EU AI Act Hits Aug 2. Here's the 5-Question Vendor Checklist Every Indian HR Team Should Ask Now.

The EU just gave HR procurement teams a €15 million reason to care about vendor architecture.

On August 2, 2026 — 102 days from today — the EU AI Act’s high-risk AI provisions become enforceable. Recruitment, CV screening, and candidate assessment are explicitly classified as “high-risk” use cases. Fines for non-compliance: up to €15 million or 3% of global annual turnover, whichever is higher.

This isn’t a doom piece. Most established AI screening vendors will pass compliance checks — they’re funded and staffed for it. But the unaudited AI is where the risk lives. And for Indian HR teams whose companies have EU parent entities, EU-based subsidiaries, or European operations, this is a live concern.

Who’s Actually In Scope

Before the checklist: let’s be clear about who needs to act.

You are in scope if:

• Your company has EU-based operations and uses AI to screen candidates for those roles
• Your parent company is EU-headquartered and you use a shared global HR platform
• You use an AI screening vendor that processes data for EU entities
• You have EU investors who may conduct due diligence on your HR practices

You may be borderline if:

• You’re a fully India-only business with no EU nexus — but you should still care about the precedent

The compliance wave is already hitting trade press. Expect a flood of content from incumbent vendors (Paradox, HireVue, Phenom) over the next 90 days — they have legal teams prepping documentation right now. If your vendor isn’t preparing for August 2, that’s a signal.

The 5-Question Vendor Checklist

These are the questions every Indian HR team should put to their current AI screening vendor — whether you’re using a global platform or a local tool. HireQwik answers yes to all five by design. Use these questions for any vendor you’re evaluating.

Question 1: Do you log every agent utterance during a candidate interaction?

What you’re asking: Is there a complete, immutable record of what the AI said to candidates during the screen? Not just the candidate’s answers — the AI’s questions, prompts, and follow-ups.

Why it matters: EU AI Act requires that high-risk AI systems maintain logs. If your vendor can’t produce a transcript of the AI’s side of the conversation, they can’t pass a compliance audit.

What good looks like: Full conversation logs, retained for a defined period, exportable on demand.

Question 2: Is there a human override before a reject decision is final?

What you’re asking: Can a human HR professional reverse, modify, or delay any AI-generated reject recommendation before a candidate is notified?

Why it matters: This is the crux of Mobley v. Workday and the heart of the EU AI Act’s “meaningful human oversight” requirement. Automated rejection without human review is the highest-risk configuration.

What good looks like: A defined workflow where no candidate receives a rejection communication until a human has reviewed the AI’s recommendation.

Question 3: Can you produce a bias audit on demand?

What you’re asking: Can the vendor run and deliver a demographic analysis of their screening outcomes — pass rates by age group, gender, language background — for your specific campaign data?

Why it matters: EU AI Act mandates bias testing for high-risk AI systems. Annual bias audits are already required under NYC Local Law 144 for any AI hiring tool used in New York City.

What good looks like: The vendor has a documented bias testing methodology, runs it on client data periodically, and can produce results within days, not months.

Question 4: Do you disclose AI use to candidates?

What you’re asking: Are candidates informed, before the interaction begins, that they are interacting with an AI system — not a human?

Why it matters: EU AI Act requires disclosure. Beyond regulation, it’s increasingly an expectation: candidates who find out post-hoc that they were talking to AI without disclosure report significantly lower employer brand perception.

What good looks like: A clear pre-screen disclosure that names AI involvement and explains what data will be collected and how it will be used.

Question 5: Is the scoring rubric inspectable?

What you’re asking: Can you see, understand, and explain to a candidate how the AI arrived at its recommendation? Is the scoring framework documented and available to HR?

Why it matters: “Black box” scoring is the central failure mode in AI hiring litigation. If you can’t explain why a candidate was classified as Strong Go versus No Go, you can’t defend the decision.

What good looks like: A documented rubric — dimensions assessed, how follow-up questions trigger, how narrative summaries are generated — that HR can read and explain.

How to Use This Checklist

Put these questions to your vendor in writing. Request documentation, not a sales call. Ask specifically for:

• Sample compliance documentation or data processing agreements referencing EU AI Act
• A sample bias audit report (anonymised from another client is fine)
• The candidate disclosure language used in your geography
• The scoring rubric or framework guide

Vendors that can’t produce these in 30 days have 90 days before the deadline. That’s not enough time to build compliance from scratch.

The Gartner Context

82% of HR leaders plan to use agentic AI by mid-2026 (Gartner). SHRM’s 2026 report found that 50%+ of TA leaders plan autonomous AI sourcing and screening within 12 months. AI adoption in HR is accelerating — which means the compliance surface area is growing faster than compliance readiness.

The vendors who will win in the next 24 months are the ones who can make a genuine claim: our architecture is defensible, our logs are complete, and our humans are in the loop. Not as a marketing line. As a documented, auditable fact.

HireQwik’s screening framework was built from day one with human-in-the-loop design: every tier reviewed by HR, full conversation logs, documented rubrics, and candidate disclosure at the start of every screen. If you want to see how that maps against the August 2 checklist, book a walkthrough and we’ll walk through it question by question.