EU AI Act and Recruitment: What's Already Mandatory Since August 2, 2026

Most recruiters think the EU AI Act is a 2027 story. Part of it already applies.

The EU AI Act is often framed as a distant deadline, set for December 2, 2027, for recruitment systems classified as "high-risk." That's true for the heaviest obligations. But part of the regulation already applies, since August 2, 2026 — and it directly concerns recruiters using a chatbot or AI-generated content anywhere in their hiring process.

What's mandatory since August 2, 2026

Article 50 of the AI Act sets out transparency obligations that have been in force since August 2, 2026:

• any chatbot embedded in an application process must be disclosed as such to candidates;
• any HR content generated by AI without human review — text, image, or video — must be identified as such.

For systems already in service on that date, a compliance grace period runs until December 2, 2026. In practice, recruitment teams already using AI tools in their candidate journey still have a few months to make sure that disclosure is properly in place — but the window is closing.

What's still ahead: recruitment classified as "high-risk"

Recruitment is listed in Annex III of the AI Act among "high-risk" use cases. The resulting obligations are more structural: traceable human oversight of AI-assisted decisions, informing candidates about the use of these systems, and a requirement for representative training data. These obligations only take effect from December 2, 2027 — but preparing now avoids a last-minute scramble, particularly around documentation and process traceability.

The penalties at stake

The regulation provides for significant financial penalties for non-compliance: up to €15 million or 3% of global annual turnover for general obligations, and up to €35 million or 7% of global annual turnover for prohibited practices.

Why this transparency requirement matters beyond the compliance checkbox

It's easy to treat these obligations as a box to tick. There's also a more direct trust issue at stake: a candidate who unknowingly talks to a chatbot, or receives an AI-generated response without human review, usually finds out one way or another — and finding out after the fact damages the relationship more than upfront disclosure would. Clearly flagging AI use isn't just a legal requirement: it's also what lets a candidate know who they're actually talking to, and keeps a human point of contact identifiable when they need one.

Frequently asked questions

Does this apply to smaller companies too? The AI Act text doesn't appear to carve out a size exemption for the Article 50 transparency obligations — a small business using a recruitment chatbot is covered the same way a large enterprise is. (To confirm precisely with legal counsel: we haven't verified whether sector- or size-based exemptions exist in the final text.)

What counts as "AI-generated content without human review"? The source material we consulted doesn't spell out an exhaustive list. As a cautious baseline, this likely covers at least automated messages sent to candidates without prior human validation (chatbot replies, auto-generated follow-up emails), and visual or video content produced by AI without human involvement. (A question for legal counsel to settle, not to interpret in marketing content.)

What if we only use an AI tool to pre-screen CVs, with no chatbot or generated content involved? That's a distinct use case, closer to the "high-risk" Annex III obligations (2027 deadline) than to Article 50. The two tracks don't fully overlap — worth keeping that distinction clear in internal team communication.

What this means in practice for recruitment teams

A few starting points, rather than waiting for the last minute:

• Map the AI tools already used in the candidate journey (chatbots, content generation, pre-screening).
• Check that every chatbot or AI-generated content use is explicitly disclosed to candidates.
• Start documenting, now, the human oversight applied to AI-assisted decisions — this traceability will be required by 2027, and building it progressively avoids a last-minute project.
• Treat the December 2, 2026 deadline as an interim milestone, not a distant date.

Clarentia's position

At Clarentia, human oversight by the recruiter stays at the center of how we design our AI agents: our tools support recruitment teams — they don't replace their decisions. That's a foundational principle, independent of the regulatory calendar — but it naturally aligns with the direction the AI Act is taking.

The takeaway

The EU AI Act isn't just a 2027 story. Part of it already applies, and the compliance window for existing systems closes on December 2, 2026. Better to treat it as a checkpoint now than a scramble in six months.