Journal

AI used to screen or rank candidates is high-risk under the EU AI Act. What is settled, what moved to 2027, and four steps to take now.

You cannot govern a hiring tool you have not named. A five-minute, five-column inventory to find every AI in your recruiting stack.

Six questions to send every AI hiring vendor, and why asking them in writing is the first page of your compliance defense under the EU AI Act.

Article 6(3) can exempt a system from high-risk status, but profiling candidates is always high-risk. How the exemption actually works for AI hiring tools.

Article 26(2) oversight means a named person with authority to overrule the tool. Four steps to build hiring oversight that survives being questioned.

Annex III point 5 puts credit scoring and insurance pricing in the high-risk tier, and adds the Article 27 FRIA. How the hiring playbook carries over.

Article 26(6) requires keeping logs at least six months. How to build a hiring AI evidence trail captured as you go, not reconstructed after the fact.

New York's bias-audit law is live now; the EU AI Act is broader and arrives in 2027. How to run one hiring compliance program that satisfies both.

Article 25 names three ways a deployer becomes a provider and inherits the heavier duties. How to stay a deployer on purpose with AI hiring tools.

Article 27's fundamental rights impact assessment does not apply to most private employers using AI hiring tools. How to tell if you are inside or outside it.

ISO/IEC 42001 is a voluntary, certifiable AI management standard. How it maps to EU AI Act deployer work, and why certification is not legal compliance.

The EU AI Act Article 4 AI literacy duty has applied since 2 February 2025, covers any AI tool not just high-risk, and needs no certificate. Three steps for HR and legal teams.

The EU AI Act's 35M EUR headline fine is the top of three tiers and rarely the one that applies to hiring. The 15M/3% middle tier, the SME cap, and the dates that matter.

The EU Commission's draft Article 6 guidelines confirm a human reviewer does not move an AI hiring tool out of high-risk. What it means, the timeline, and what to do.

Article 50's transparency duty applies from 2 August 2026 and was just reaffirmed. When a candidate deals with your AI, they have to be told. Who it binds, the narrow 'obvious' exception, and what to do.

NYC's AEDT law has been enforced since 2023. The bias audit must be independent and no more than a year old, the summary published, and candidates given 10 business days notice. How to find your gaps.

Article 5 bans AI that infers emotions of workers and candidates, in force since 2 February 2025. Who it reaches, the narrow medical or safety exception, and what to check in your hiring stack.

The Commission's draft Article 6 guidelines on classifying high-risk AI are open for comment until 23 July 2026. Why a human in the loop is not the off-ramp, why profiling is a hard stop, and what to inventory now.

Article 10 sets a data governance checklist for high-risk hiring AI. What to actually demand from your vendor, and why it is your risk even though they are the provider.

Before a high-risk AI system goes live at work, employers must inform affected workers and their representatives first. What the duty covers, who it lands on, and how to build it into rollout.