Insights & perspectives on modern recruitment

Let's get the headline out of the way: the EU AI Act's compliance clock for high-risk employment AI has moved. The original date of 2 August 2026 for the heaviest obligations was always circled in red on every recruitment technology team's calendar. Then the European Commission proposed a deferral, and after a series of trilogue negotiations,

the European Parliament, the Council of the EU, and the Commission reached political agreement on 7 May 2026, shifting the deadline for high-risk AI in employment from 2 August 2026 to 2 December 2027.

Cue the sighs of relief across talent acquisition teams everywhere. And then, almost immediately, the misreading of what that delay actually means.

What Got Deferred and What Didn't

The deferral is real and it matters.

The Act classifies AI systems used in employment-related decisions as high-risk, including tools used for recruitment, candidate selection, performance evaluation, task allocation, monitoring of workers, and decisions on promotion or termination.

The full conformity assessment regime, the quality management documentation, the formal bias auditing requirements tied to those Annex III systems: all of that now has until December 2027.

But here's where the collective sigh of relief becomes a trap.

The ban on emotion-recognition AI in the workplace has been in force since February 2025, and it is not a "manage the risk" category. It is an absolute prohibition: an employer may not use an AI system that infers the emotions of employees or candidates, for example by analyzing facial expressions or tone of voice.

If your hiring stack includes any such tool, you are already out of compliance, today, regardless of any Omnibus deferral.

More directly relevant to day-to-day recruiting operations:

most of the transparency duties under Article 50 take effect on 2 August 2026, exactly as originally planned. This includes the duty to tell candidates they are interacting with an AI system, for example an AI chatbot in a recruitment process.

That specific obligation did not move. It lands this week.

And there is the AI literacy obligation to consider.

The AI literacy requirement has been in force since February 2025, requiring employers to ensure that staff who use AI systems, or whose work is affected by them, receive adequate training, documented rather than just verbal guidance.

So the picture is actually three separate layers: prohibited practices that have been in force for over a year, transparency duties arriving now, and heavy conformity documentation pushed to 2027. Conflating them is how organisations end up both complacent and caught out at the same time.

Why Recruiters Are Deployers, Not Bystanders

One of the most consequential misunderstandings in the industry right now is the assumption that compliance sits with the vendor. Buy a tool, let the vendor worry about the paperwork.

Many organizations believe that if they purchase an AI tool from a vendor, the responsibility for legal compliance rests with the provider. The EU AI Act directly challenges this assumption by creating a chain of shared liability. The law distinguishes between "providers," the companies that build AI systems, and "deployers," the businesses that use them. If your company uses AI to screen résumés or assess candidates, you are a deployer and share responsibility for the system's fairness and transparency.

This is the structural shift that the deferral does not change. Recruiters and hiring teams who use AI tooling for any part of the selection process have obligations that are theirs to own, not outsource.

Deployers must implement human oversight mechanisms and retain automated logs for at least six months.

That is an operational requirement, not a legal technicality. It means your process needs to be designed with audit trails in mind from the start.

A recruiter based in London or New York screening candidates for EU-based roles falls within scope, so global staffing groups should apply one EU-grade compliance standard across their hiring stack.

The regulation does not stop at European borders if the output of the hiring process touches candidates based in the EU.

The Delay Is Runway, Not Permission

The single most useful reframe here is this:

the later deadline is added runway to audit your AI tools, establish governance frameworks, and implement the necessary safeguards. It is not a reason to defer the work.

The organisations that treat December 2027 as a start date rather than a deadline will face exactly the same scramble that played out before August 2026, only with steeper penalties looming.

Enterprise compliance programs already lag behind the scale of AI deployment, with over half of organizations lacking systematic AI inventories, and harmonized technical standards arriving eight months late, compressing implementation timelines further.

The practical implication for recruitment teams is that the audit work is not optional or theoretical. It is the foundation on which every other compliance step rests, and it takes longer than anyone estimates. If you do not know precisely which tools in your hiring stack are doing any form of candidate scoring, ranking, or filtering, you cannot assess your risk exposure. You cannot brief your vendors. You cannot implement meaningful human oversight. The inventory is step zero, and many teams have not taken it.

What Transparency Looks Like in Practice

With transparency duties live right now, the most immediate operational question for any recruiter using AI in their process is: do candidates know? Not in the abstract, buried-in-a-privacy-policy sense. In the active, clear, point-of-interaction sense.

Candidates often feel uneasy about the impersonal nature of automation, but research from Arup found success by openly communicating the benefits of automation. Candidates want to know that automation is not being used to weed them out, but rather to bring them in for human conversations. Proactively framing technology as an enabler rather than a barrier can maintain a personal touch even while managing high application volumes.

This is both a compliance point and a candidate experience point, and the fact that they align is not a coincidence. The legislation is, at its core, codifying what good practice already looks like. Recruiters who have been thoughtful about how they present AI-assisted processes to candidates are not starting from zero. They are formalising what they already do.

The harder version of transparency is about what happens when a candidate is not progressed.

The EU AI Act reflects a legislative view that these systems have a direct effect on access to employment and on fundamental rights, so the question for recruiters is not "am I concerned?" but "how do I document and govern tools that I must presume to be high-risk?"

That shift in framing, from compliance as a checkbox to governance as a design principle, is where the regulation has its most lasting effect on how recruitment technology gets built and used.

Tools like those built into the Floats platform are designed with this accountability logic already embedded: every AI-assisted action is visible, auditable, and keeps the recruiter in the decision seat. That is not an accident. It is what responsible deployment looks like in practice.

The Honest Assessment

The EU AI Act is not a bureaucratic inconvenience that can be deferred indefinitely.

The delay is not a repeal. Employment-related AI remains squarely in the Act's high-risk category when used for recruitment, selection, promotion, termination, task allocation, or performance monitoring.

The category did not change. The timeline moved.

What that creates is a specific window of opportunity for recruitment leaders who want to get ahead of the curve rather than scramble to meet it. Audit your tools now, while there is time to make considered decisions. Brief your vendors on their obligations as providers. Build the documentation habit before it becomes a crisis. Train your team on what AI literacy actually means for their daily workflow, not as a one-off presentation, but as a genuine shift in how they understand the tools they use.

The deferral bought time. How that time gets used will define which recruitment operations are genuinely ready when December 2027 arrives, and which are filing for extensions.