The EU AI Act and Employment: What the Delayed High-Risk Deadlines Mean for Employers and Their Lawyers

The EU AI Act classifies a specific set of employment-related uses of artificial intelligence as "high-risk" — including AI used in recruitment and candidate selection, decisions on promotion and termination, task allocation, and performance monitoring — triggering a distinct set of compliance obligations for the organisations deploying these systems. The high-risk obligations for this category were originally due to apply from August 2026. They have since been delayed, a change solicitors advising employer clients on AI compliance need to understand precisely, since the delay changes timing but not the underlying substance of the obligations.

This guide sets out what has actually changed, why the delay is not the same as a repeal, and what employer clients should still be doing now.

What has changed: the revised timeline

The compliance deadlines for high-risk AI systems under the EU AI Act have been delayed from the original August 2026 date to two later dates depending on the type of system involved: 2 December 2027 for stand-alone high-risk AI systems, and 2 August 2028 for high-risk AI that is embedded within regulated products. This distinction matters in advising clients — a business using a discrete, purpose-built AI recruitment screening tool falls into the stand-alone category with the earlier of the two revised deadlines, while AI functionality embedded within a broader product, such as certain categories of workplace safety equipment or machinery with integrated AI monitoring, falls into the later, embedded-systems deadline.

Employment-related AI uses — recruitment and candidate selection, promotion decisions, termination decisions, task allocation, and performance or behaviour monitoring — remain squarely within the Act's high-risk category throughout this delay. Nothing about the underlying classification of these use cases has changed; only the date by which the associated compliance obligations must be met has moved.

Why "delay" does not mean "repeal"

It would be a significant misreading of this development to advise a client that employment AI compliance is now off the table until 2027 or 2028 in any meaningful sense. The delay addresses implementation timing — recognising that both regulators and businesses needed more time to build the practical infrastructure (conformity assessment processes, technical documentation standards, notified body capacity) required to enforce and comply with the high-risk regime properly. It is explicitly not a substantive change to the underlying obligations or the classification of employment AI as high-risk.

For solicitors advising employer clients, this distinction should shape the advice given: clients should not treat the delay as a reason to deprioritise AI governance work, but rather as additional runway to get the underlying compliance architecture right before the enforceable deadline arrives. Clients who treat the delay as removing the issue entirely risk having far too little runway when the December 2027 deadline for stand-alone systems approaches.

What high-risk classification will require

Once the relevant deadline takes effect, deployers of high-risk employment AI systems will need to meet obligations that typically include maintaining adequate technical documentation about the system, ensuring appropriate human oversight of automated decisions, monitoring the system's performance in the specific context of its actual use, and in many cases informing affected individuals — such as job applicants or employees subject to performance monitoring — that a high-risk AI system is being used in relation to them. Employers will also need clear internal processes for handling situations where an automated decision is challenged, since human oversight requirements are a central feature of the high-risk framework rather than an optional addition.

These substantive requirements connect closely to the wider set of practical AI governance issues covered in our guide to AI governance for managers, since much of what the Act requires in substance — clear oversight, documented decision processes, and transparency with affected individuals — overlaps with what good AI governance practice already recommends regardless of the specific regulatory deadline.

What employer clients should be doing now

• Inventory AI use in employment decisions. Clients should identify every system used in recruitment, promotion, termination, task allocation, or performance monitoring that involves any degree of AI or automated decision-making, since this inventory is the starting point for any high-risk compliance assessment.
• Build documentation and oversight processes ahead of the deadline. Rather than waiting until closer to December 2027, clients should use the extended timeline to build technical documentation and human oversight processes progressively, reducing the compliance burden concentrated near the deadline.
• Monitor further regulatory guidance. As the revised deadlines approach, further implementing guidance and standards are likely to be published; advising clients to monitor developments rather than relying solely on the Act's text is prudent given how much of the practical detail is still being worked out.

Frequently asked questions

What was the original deadline for high-risk AI obligations under the EU AI Act?

The high-risk obligations, including those covering employment-related AI, were originally due to apply from August 2026.

What are the revised deadlines?

The revised deadlines are 2 December 2027 for stand-alone high-risk AI systems and 2 August 2028 for high-risk AI embedded within regulated products.

Does the delay mean employment AI is no longer classified as high-risk?

No — employment-related AI uses such as recruitment, promotion, termination, task allocation, and performance monitoring remain within the high-risk category; only the compliance deadline has moved, not the underlying classification.

The delayed EU AI Act deadlines give employers genuine additional time to prepare, but solicitors advising on this area should be precise in distinguishing a timing delay from a substantive change — the underlying obligations for high-risk employment AI are unchanged and clients should be encouraged to use the extra time productively rather than treat the issue as resolved. Learnsignal's CPD courses cover AI regulation and its practical implications for legal practice.