EU AI Act Delay for HR and Hiring AI: The New Timeline and What to Do Now
The Digital Omnibus moved EU AI Act rules for AI hiring and HR tools to December 2, 2027. Here is the confirmed timeline and what HR teams should do now.

If your company uses AI to screen resumes, rank candidates, schedule shifts or score performance, the EU AI Act deadline of August 2, 2026 no longer applies to those tools. The EU's Digital Omnibus on AI moved it to December 2, 2027, and that change is now law, not a proposal. This guide explains what was delayed, what was not, and gives a step by step plan for small and mid sized HR teams, including US employers with candidates or staff in Europe. It reflects official EU sources checked on November 8, 2026, and was updated on November 18, 2026 with findings from Littler's 2026 European Employer Survey on how employers are preparing.
The short answer: what changed for HR AI and what did not
The AI Act (Regulation (EU) 2024/1689) sorts AI systems by risk. AI used in employment is listed as "high risk," which brings the heaviest obligations short of an outright ban. Under the original text, most high risk rules were due to apply on August 2, 2026.
Regulation (EU) 2026/1744, known as the Digital Omnibus on AI, amended Article 113 of the AI Act. Chapter III, Sections 1 to 3 (the core high risk requirements and the obligations of providers and deployers) now apply:
- from December 2, 2027 for AI systems that are high risk because they appear in Annex III, which includes employment and worker management; and
- from August 2, 2028 for AI that is high risk because it is a safety component of a regulated product listed in Annex I, such as machinery, toys or lifts.
What did not change: employment AI is still high risk, the list of covered HR uses is the same, the prohibitions that started in February 2025 still apply, and employers still carry a (now softer) AI literacy duty. The delay is a postponement, not a repeal. The obligations simply start 16 months later than originally planned, which leaves a little over a year from mid November 2026.
Which HR and hiring tools count as high risk
Annex III, point 4 of the AI Act, titled "Employment, workers' management and access to self-employment," covers two groups of AI systems.
Recruitment and selection. AI intended to be used to recruit or select people, in particular to:
- place targeted job advertisements;
- analyze and filter job applications; and
- evaluate candidates.
Decisions about people already working for you. AI intended to be used to:
- make decisions affecting the terms of work relationships;
- decide on promotion or termination;
- allocate tasks based on individual behavior or personal traits; and
- monitor and evaluate the performance and behavior of workers.
In practice, that can capture resume screening and ranking in an applicant tracking system, AI interview scoring, screening chatbots, job ad targeting, AI driven shift or task assignment, and performance analytics. The Commission's own example for this category is "CV-sorting software for recruitment."
Not every AI feature in an HR suite is high risk. A tool that drafts a job description does not filter, evaluate or decide about people. The test is the system's intended purpose and your actual use. Providers can also document under Article 6(3) that an Annex III system does not pose a significant risk, so ask vendors whether they rely on that exception and request their written assessment.
If you are still mapping which of your systems are HR, HCM or payroll tools, our guide to HRIS vs HCM vs payroll software is a useful starting point.
The new EU AI Act timeline for HR teams
Dates below come from the AI Act as amended by Regulation (EU) 2026/1744.
| Date | What applies | Status on November 8, 2026 | Why HR teams care |
|---|---|---|---|
| August 1, 2024 | AI Act enters into force | Law | Starting point for all deadlines |
| February 2, 2025 | Prohibited practices (Article 5), original AI literacy duty (Article 4) | Applying | Emotion recognition of workers is banned |
| August 2, 2025 | Governance rules and general purpose AI model rules | Applying | Mostly affects model providers, not employers |
| July 27, 2026 | Digital Omnibus on AI enters into force | Law | Softer AI literacy wording, new dates locked in |
| August 2, 2026 | General date of application, including most transparency rules and national enforcement structures | Applying | Authorities are now active; non high risk rules apply |
| December 2, 2026 | New ban on AI that generates non consensual intimate content; grace period ends for marking AI generated content | Upcoming, law | Limited direct HR impact |
| December 2, 2027 | High risk obligations for Annex III systems, including employment (Annex III, point 4) | Upcoming, law | Deployer duties under Article 26 start |
| August 2, 2028 | High risk obligations for AI in regulated products (Annex I) | Upcoming, law | Relevant if you use AI enabled machinery |
From proposal to law: how the delay moved through Brussels
Early coverage called the delay "proposed," then "agreed." Here is each stage.
- November 19, 2025: Commission proposal. The Commission's Digital Omnibus on AI (COM(2025) 836) did not set a fixed date. It would have started Annex III rules six months after a Commission decision confirming standards and guidance were ready, with December 2, 2027 as the latest date. It also would have turned the AI literacy duty into an obligation on the Commission and Member States to "encourage" companies.
- May 7, 2026: provisional political agreement. Parliament and Council negotiators reached a deal. The text was agreed politically but not yet law.
- June 16, 2026: Parliament vote. Parliament approved the text with 423 votes in favour, 57 against and 174 abstentions.
- June 29, 2026: Council adoption. The Council gave its final green light, confirming a fixed timeline instead of the proposal's flexible mechanism: December 2, 2027 for stand alone high risk systems and August 2, 2028 for systems embedded in products.
- July 24 to 27, 2026: publication and entry into force. Regulation (EU) 2026/1744 was published in the Official Journal on July 24 and entered into force on July 27, 2026, days before the original deadline.
What is confirmed today: the December 2, 2027 date for employment AI is final and in force. What is still open: guidance and harmonised standards for high risk systems are still arriving (the Commission reports the first harmonised standard supporting the AI Act was published on October 19, 2026), and national authorities will develop their own enforcement practice. Nothing in the official record points to another postponement, so do not plan on one.
What European employers are doing now: Littler's 2026 survey
The delay has not slowed AI adoption at work. Employment law firm Littler's ninth annual European Employer Survey, published in November 2026, gathered responses from 498 executives, in-house lawyers and HR leaders across 13 European countries through an online survey fielded in August and September 2026. Its headline findings for HR teams:
- AI is already in HR. 94% of respondents report using AI in HR or employment law functions.
- AI is reshaping headcount. The share of employers that made workforce reductions because of actual or anticipated AI use rose from 13% in 2025 to 28% in 2026, and the share that reduced hiring rose from 13% to 32%.
- Regulation is a top concern. 67% expect AI related regulatory changes to affect workforce management in the next 12 months, up from 58% in 2025.
- Employees use AI in disputes too. 74% report a rise in employees using AI to prepare or support complaints, grievances or legal claims, and 49% of that group use AI tools themselves to review or respond.
Preparation for the AI Act is broad but shallow. 93% had taken at least one step, but no single step had been taken by a majority:
| Preparation step | 2026 | 2025 |
|---|---|---|
| Reviewing or updating internal policies on AI use | 49% | 51% |
| Identifying which compliance obligations may apply | 49% | 47% |
| Internal audits or risk assessments of AI use | 46% | 34% |
| Training for relevant teams | 38% | 40% |
| Creating or updating an inventory of AI tools | 36% | 36% |
| Engaging external advisers | 26% | 18% |
| Assigning internal ownership for compliance | 25% | 29% |
| No steps taken | 7% | 1% |
Source: Littler, 2026 European Employer Survey Report. Respondents could select more than one step.
What HR teams should do while the rules are uncertain
The survey points to a clear gap: only a quarter of employers have someone accountable for AI Act compliance, and only about a third have an inventory of the AI tools they use. Those are the two steps everything else depends on, and they cost little. A practical order for SMB HR teams:
- Name one owner now. Give a named person in HR, legal or operations responsibility for AI in hiring and people management, with time to do it.
- Build the AI inventory before anything else. You cannot classify, document or explain tools you have not listed (Step 1 below).
- Do not wait for final guidance on what is certain. The emotion recognition ban and the AI literacy duty already apply, and the high risk date of December 2, 2027 is law.
- Plan for AI in disputes. Expect more detailed, AI drafted complaints and data requests. Make sure managers document decisions, and that anyone using AI to respond checks every output.
- Treat workforce reductions linked to AI carefully. Decisions on termination supported by AI fall in the high risk category from December 2027, and national consultation rules already apply today.
- Track the other delayed law. Littler found 84% of respondents say uneven national implementation of the EU Pay Transparency Directive has complicated compliance planning. Plan both programs together, since both touch HR data and pay decisions.
If you need a starting point for internal rules, our guide on how to write an AI use policy for work covers approved tools, data handling and human review.
What already applies to HR teams today
Several AI Act provisions affect employers right now, regardless of the high risk delay.
The ban on emotion recognition at work
Since February 2, 2025, Article 5 has prohibited placing on the market, putting into service or using AI systems "to infer emotions of a natural person in the areas of workplace and education institutions," except for medical or safety reasons. That can catch video interview tools that claim to read enthusiasm, honesty or stress from facial expressions or voice, and call center tools that score agents' emotional state. Violations of Article 5 carry the Act's highest fines: up to EUR 35 million or 7% of worldwide annual turnover, whichever is higher. If you use any tool marketed with "sentiment," "emotion" or "affect" analysis of candidates or employees, review it now.
AI literacy, in its softened form
The original Article 4 required providers and deployers to "ensure, to their best extent, a sufficient level of AI literacy" of their staff. The Omnibus replaced it. Article 4 now says providers and deployers "shall take measures to support the development of AI literacy" of staff and others operating or using AI systems on their behalf, taking into account their knowledge, experience and the context of use. It adds that this "does not require providers or deployers to guarantee any specific level of AI literacy of any individual." The Commission must publish practical examples of compliance and support SMEs.
Note what happened here: the Commission had proposed removing the company duty entirely, but the final law kept a lighter one. For HR, that means recruiters and managers using AI tools should get practical training, and you should keep a simple record of it.
Existing data protection law
The AI Act sits on top of EU data protection law, not in place of it. Article 26(9) points deployers to their duty to carry out a data protection impact assessment under Article 35 of the GDPR, using information the provider supplies.
What employers will owe from December 2, 2027
When the high risk rules apply, an employer using an Annex III HR tool is a deployer. Article 26 of the AI Act lists the core deployer obligations. In plain terms, you will need to:
- Use the system as instructed. Take technical and organizational measures to use the tool according to the provider's instructions for use.
- Assign human oversight. Give oversight to people with "the necessary competence, training and authority," and support them.
- Control input data. Where you control the input data, make sure it is relevant and sufficiently representative for the intended purpose.
- Monitor and report. Watch how the system performs. If you think it presents a risk, inform the provider and the market surveillance authority and suspend use. Report serious incidents.
- Keep logs. Keep the automatically generated logs under your control for an appropriate period of at least six months, unless other law says otherwise.
- Inform workers first. Before using a high risk system at the workplace, employers must inform workers' representatives and the affected workers.
- Inform candidates and others. Deployers of Annex III systems that make or assist decisions about people must tell those people they are subject to the system.
- Explain decisions on request. Under Article 86, a person affected by a decision based on a high risk system's output that significantly affects them can ask the deployer for "clear and meaningful explanations" of the AI's role and the main elements of the decision.
Fines for breaching deployer obligations under Article 26 reach up to EUR 15 million or 3% of worldwide annual turnover, whichever is higher. For SMEs, including start ups, each fine is capped at whichever of the two amounts is lower.
A note on tools already in use
Article 111(2), as amended, says the Act applies to high risk systems placed on the market or put into service before December 2, 2027 only if their design changes significantly from that date. For cloud HR software that is updated constantly, relying on this is risky. Plan to comply rather than to qualify for an exception.
When an employer becomes the provider
Most SMBs buy HR AI, so they are deployers. But under Article 25 a deployer that puts its own name or trademark on a high risk system takes on provider obligations, which are much heavier. If you build or rebrand your own screening model, get legal advice.
What US employers with EU candidates or staff need to know
The AI Act reaches beyond EU borders. Under Article 2, it applies to providers and deployers located in a third country "where the output produced by the AI system is used in the Union." A US company that uses an AI screening tool to evaluate applicants for a role in Ireland, or that monitors the performance of employees based in Germany, should assume the deployer obligations apply to that use.
Practical implications for US businesses:
- Map by where people are, not where servers are. The question is whether candidates or workers in the EU are affected.
- Separate workflows if needed. You can run EU hiring with AI ranking turned off or with documented human review. Make that choice deliberately and record it.
- Expect national labor rules on top. Article 26(7) says worker information must follow EU and national law and practice on informing workers and their representatives, which can include works council consultation.
- Keep US rules in view. US state and local rules on automated hiring tools run on their own timelines, so one aligned program is usually cheaper than two.
Step 1: Inventory every AI feature that touches candidates or employees
List each feature, not just each vendor: ATS, video interviews, assessments, job ad distribution, HRIS analytics, scheduling and productivity monitoring. Note what it does, which decisions it informs, whether EU based people are affected, and who owns it. Features turned on by default count. A simple spreadsheet is enough.
Step 2: Classify each use against Annex III and Article 5
Mark each feature as likely high risk (filters, ranks, evaluates or decides about people under Annex III, point 4), possibly prohibited (infers emotions of workers or candidates), or lower risk (drafting, summarizing, search). Stop using anything that infers emotions unless it is clearly for medical or safety reasons. Flag uncertain cases for legal review.
Step 3: Ask vendors for their AI Act documentation and roadmap
Your vendors, as providers, carry the heaviest obligations, and you will depend on their instructions for use to comply with Article 26. Send a short questionnaire and add it to your software vendor evaluation and RFP process. Useful questions:
| Question for the vendor | Why it matters |
|---|---|
| Do you consider this feature high risk under Annex III, point 4? If not, where is your Article 6(3) assessment? | Tells you which obligations apply |
| What is your plan to complete conformity assessment and registration before December 2, 2027? | Shows whether the product will be legally usable in the EU |
| Will you provide instructions for use covering human oversight, accuracy and known limits? | You must follow these under Article 26(1) |
| What logs does the system generate, who controls them and how long are they kept? | You must keep logs you control for at least six months |
| How do you test for and correct bias, and can you share results? | Supports your data protection assessment and fairness checks |
| Does any feature infer emotions, personality or honesty from video, voice or text? | Possible Article 5 prohibition |
| Can EU workflows run with AI ranking off or with mandatory human review? | Gives you a fallback if a feature is not ready |
Step 4: Put AI Act terms into contracts and renewals
Many HR software contracts signed this year will still run in December 2027. Add clauses covering AI Act compliance as provider, instructions for use, notice of significant changes, incident cooperation, log access, and a penalty free exit if a feature cannot lawfully be used. Our SaaS contract checklist covers the standard terms to pair with these.
Step 5: Design real human oversight
Decide who reviews AI outputs at each stage and what they can override. A recruiter who cannot reject a ranking is not oversight. Write a one page procedure per tool: what the AI does, what the human checks, and when to escalate or suspend use.
Step 6: Draft notices for candidates, workers and worker representatives
Prepare plain language notices saying an AI system is used, for what purpose, and how to ask for human review or an explanation. Put candidate notices in the application flow. For employees, inform them and their representatives before rollout, and consult where national law requires it. Our guide on how to roll out new software to employees has a communication plan you can adapt.
Step 7: Train the people who use AI tools
Give recruiters and hiring managers short, role specific training: how the tool works, its limits, how to spot errors or bias, how to override it, and the emotion recognition ban. Record who was trained and when. This supports Article 4 today and Article 26 oversight from December 2027.
Step 8: Set up monitoring, logging and a review cadence
Confirm you can keep logs for at least six months. Assign an owner to review outcomes, such as pass rates by stage, every quarter. Write down what triggers suspension and who contacts the vendor and the authority. Revisit the inventory twice a year, since vendors keep adding AI features.
How this affects your next HR software purchase
The delay is a buying advantage. You now have time to choose vendors that can show credible AI Act readiness instead of rushing to fix tools already in place. When you compare platforms, treat AI Act documentation, human review controls and log access as selection criteria, alongside price and usability. Our guide on how to choose HR software for a small business walks through a full evaluation, and you can browse vendors in our HR and payroll software category.
This guide is general information for business software buyers, not legal advice. AI Act obligations depend on your specific use, location and national law, so consult qualified counsel for your situation.
Frequently asked questions
Did the EU cancel the AI Act rules for AI hiring tools?
No. The Digital Omnibus on AI only moved the date. AI used to place targeted job ads, filter applications, evaluate candidates, or make decisions on promotion, termination, task allocation and performance monitoring stays in the high risk category under Annex III, point 4. Those obligations now apply from December 2, 2027.
Is the December 2, 2027 date final or still a proposal?
It is final law. Parliament approved the text on June 16, 2026, the Council gave its final approval on June 29, 2026, it was published in the Official Journal on July 24, 2026 as Regulation (EU) 2026/1744 and it entered into force on July 27, 2026. Only a new amendment could change it again.
Does the EU AI Act apply to a US company hiring in Europe?
It can. The Act covers providers and deployers located outside the EU where the output of the AI system is used in the EU. A US employer that screens EU based candidates or monitors EU based staff with AI should plan as if the deployer obligations apply to it.
Is AI literacy training still mandatory under the EU AI Act?
A softer version is. Since the Omnibus, Article 4 requires providers and deployers to take measures to support the AI literacy of staff who operate or use AI systems on their behalf. It no longer requires a guaranteed level of literacy for each individual, but you should still be able to show what you did.
Who is responsible for compliance, the HR software vendor or the employer?
Both, for different things. The vendor, as provider, handles conformity assessment, technical documentation and instructions for use. The employer, as deployer, must use the system according to those instructions, assign trained human oversight, monitor it, keep logs it controls for at least six months and inform workers and candidates.
Are European employers ready for the EU AI Act rules on HR AI?
Mostly not yet. In Littler's 2026 European Employer Survey of 498 employers, fielded in August and September 2026, 93% had taken at least one step to prepare, but no single step had been taken by a majority. 49% had reviewed AI policies, 46% had audited AI use, 36% had an inventory of AI tools and only 25% had assigned internal ownership for compliance.
Sources
- Regulation (EU) 2026/1744, Digital Omnibus on AI, EUR-Lex (checked November 8, 2026)
- Regulation (EU) 2024/1689, the AI Act, EUR-Lex (checked November 8, 2026)
- Commission proposal COM(2025) 836, Digital Omnibus on AI, EUR-Lex (checked November 8, 2026)
- European Commission, AI Act policy page (checked November 8, 2026)
- European Commission, AI Omnibus enters into force (checked November 8, 2026)
- Council of the EU press release, 29 June 2026 (checked November 8, 2026)
- European Parliament press release, 16 June 2026 (checked November 8, 2026)
- Littler: 2026 European Employer Survey Report, November 2026 (checked November 18, 2026)
- Personnel Today: European firms grapple with regulatory uncertainty amid AI takeup, November 18, 2026 (checked November 18, 2026)