The EU AI Act Got Delayed, Except for the Parts That Were Not: What It Means If You Build HR Software
• 3 min read • Ahmed Daoudi
For a year, August 2, 2026 was circled on every compliance calendar as the day the EU AI Act's high-risk rules would bite. Then the EU passed the Digital Omnibus on AI and moved the deadline. Many teams read the headline, relaxed, and missed the details.
What changed, and what did not
The Omnibus is Regulation (EU) 2026/1744 and entered into force on July 27, 2026. Its main effects:
- High-risk obligations for stand-alone systems listed in Annex III move to December 2, 2027. Annex III includes employment and worker management, creditworthiness, education and access to essential services.
- High-risk AI embedded in regulated products (Annex I, such as medical devices and machinery) moves to August 2, 2028.
- Machine-readable labelling of AI-generated content for systems already on the market before August 2, 2026 moves to December 2, 2026.
What stayed put: the Act's general application date remained August 2, 2026. Transparency obligations under Article 50, for example telling people they are talking to a chatbot and disclosing deepfakes, apply now. The Commission's AI Office also gained its enforcement powers over general-purpose AI models on that date. The European Parliament used the same package to ban so-called "nudifier" apps.
Why HR software developers should care most
I build HR and payroll systems on Odoo, so this one is personal. Employment is explicitly a high-risk area. If your product uses AI to screen CVs, rank candidates, evaluate performance, allocate tasks or inform promotion and termination decisions, it is likely a high-risk system once the rules apply. That brings obligations around risk management, data quality, logging, human oversight and documentation.
The law also reaches beyond EU borders. A company in Morocco or anywhere else whose system is used for EU employees is in scope. Many of us building ERP and HR tools serve European clients, so "we are not in the EU" is not an answer.
How to use the extra sixteen months
- Inventory your AI features. List every place a model scores, ranks or recommends something about a person.
- Log decisions now. Store inputs, model version and outputs for each automated recommendation. Retrofitting logging later is painful.
- Design human oversight into the UI. A recruiter should see why a candidate was ranked low and be able to override it, with the override recorded.
- Document training data: where it came from, what it represents, which groups are under-represented.
- Ship the transparency pieces today. If your HR portal has a chatbot, it must say it is an AI. That obligation is already live.
My take
The delay is a gift only if you use it. Teams that treat December 2027 as the start date will be scrambling in late 2027, the same way many scrambled before GDPR. Teams that build logging and oversight into the product now will have a selling point, because enterprise buyers will ask for it long before the regulator does.
This article is a developer's summary, not legal advice. Check specific obligations with a qualified lawyer.