The EU Digital Omnibus Package has changed the timeline for certain EU AI Act obligations, but not their overall direction. In this episode, L+E Global colleagues Luís Neto Galvão and Éric Le Quellenec offer five practical steps - inventory and classify AI-enabled HR tools, strengthen oversight and vendor contracts, mitigate bias, train relevant teams, and prepare recruitment practices for compliance - that GCs and HR leaders at multinational businesses should take now to fortify their employment decision-making AI programs.
Audio and video play straight from the show's own hosting. PodSubs stores and serves no media.
Transcript
From the show. Tap a time to play from there.
NETO GALVÃO: If our clients haven't started, NETO GALVÃO: now is the time to put a compliance plan in place. NETO GALVÃO: A year goes very quickly when you have to map your tools, NETO GALVÃO: renegotiate vendor contracts, and train your teams. LE QUELLENEC: Five words: map, govern, debias, train, frame. LE QUELLENEC: Map your HR AI tools, starting from the GDPR record. LE QUELLENEC: Govern them through rules, contracts, and the ability LE QUELLENEC: to switch them off. Tackle bias, including in prompts.
LE QUELLENEC: Train your people and keep proof. Frame recruitment uses precisely, LE QUELLENEC: because that is where the risks spike first. HOST: Welcome to the L&E Global Data Privacy and AI Podcast Series, HOST: where we discuss privacy, AI, and the workplace. HOST: Today's topic is AI and human resources after the EU AI Act Omnibus, HOST: and how to build your AI Act compliance program.
HOST: We're joined today by Eric Le Quellenec of Flichy HOST: Grangé Avocats in France and Luís Neto Galvão of SRS Legal in Portugal. HOST: Thank you both for being here, Eric and Luís. NETO GALVÃO: Thank you, Mary, and welcome to this podcast. NETO GALVÃO: Today, we're talking about artificial intelligence in human resources NETO GALVÃO: and about a legal text that has just changed the timetable: NETO GALVÃO: the Digital Omnibus. Before we get into the practical side, NETO GALVÃO: can you remind us,
NETO GALVÃO: Eric, what exactly happened and what has and hasn't been postponed? LE QUELLENEC: Thank you, Luís. Happy to be here with you. LE QUELLENEC: Back in November 2025, the Commission presented the Digital Omnibus, LE QUELLENEC: a package to simplify European Union (EU) digital law. LE QUELLENEC: It has two parts. LE QUELLENEC: The first reopens the General Data Protection Regulation (GDPR). LE QUELLENEC: It is still being negotiated, so we'll leave it aside today. LE QUELLENEC: The second amends the AI Act, and that one is now law.
LE QUELLENEC: It entered into force on July 27, 2026. LE QUELLENEC: The headline change is the delay for high-risk systems. LE QUELLENEC: The obligations for the systems listed in Annex III, LE QUELLENEC: including recruitment and career management, LE QUELLENEC: were due to apply in August of this year. LE QUELLENEC: They now apply from December 2, 2027, LE QUELLENEC: and from August 2, 2028, for AI built into regulated products.
LE QUELLENEC: Not everything has been postponed. LE QUELLENEC: The prohibitions have applied since February 2025, LE QUELLENEC: the AI literacy obligations as well, LE QUELLENEC: and the transparency rules since August 2026. LE QUELLENEC: Clients should not think they have nothing to do until 2027. NETO GALVÃO: That's really our message today, Eric. NETO GALVÃO: If our clients haven't started, NETO GALVÃO: now is the time to put a compliance plan in place. NETO GALVÃO: A year goes very quickly when you have to map your tools,
NETO GALVÃO: renegotiate vendor contracts, and train your teams. NETO GALVÃO: Let's go through the main steps with concrete examples. NETO GALVÃO: Regarding the obligation to map and classify, NETO GALVÃO: the first step is obviously to know what you have. NETO GALVÃO: Where should an employer start, and how does it work NETO GALVÃO: out whether the tools are actually caught by the AI Act? NETO GALVÃO: Employers should have an advantage NETO GALVÃO: because they often forget that their GDPR
NETO GALVÃO: record of data processing activities is a very useful tool. NETO GALVÃO: Recruitment, performance reviews, training, and internal mobility NETO GALVÃO: are all already listed, and that is the natural starting point. NETO GALVÃO: Then, for each line, there are two questions: NETO GALVÃO: Is there AI involved? If so, what does it actually do? LE QUELLENEC: On the first question, the key criterion is inference. LE QUELLENEC: The system infers from its inputs how to produce outputs,
LE QUELLENEC: predictions, recommendations, or decisions, LE QUELLENEC: rather than simply applying rules someone wrote. A spreadsheet LE QUELLENEC: that filters candidates by years of experience is not AI. LE QUELLENEC: A tool that learns from past hiring decisions to rank applicants is. LE QUELLENEC: Then comes the risk level, and three levels matter for HR. LE QUELLENEC: First, prohibited practices under the AI Act. For employers, LE QUELLENEC: the key one is emotion recognition in the workplace,
LE QUELLENEC: save for medical or safety reasons. LE QUELLENEC: A video interview that scores a candidate's LE QUELLENEC: enthusiasm from their face is not high risk; it is simply prohibited. LE QUELLENEC: Some of those tools are still on the market. LE QUELLENEC: Second, high-risk systems. Point 4 of Annex III covers two families: LE QUELLENEC: recruitment, including targeting job ads, LE QUELLENEC: filtering applications, and evaluating candidates; LE QUELLENEC: and decisions on the employment relationship,
LE QUELLENEC: including promotion, LE QUELLENEC: termination, task allocation based on behavior or personal traits, LE QUELLENEC: and monitoring and evaluating performance. LE QUELLENEC: Third, the exception. A listed system is not high risk LE QUELLENEC: if it performs a narrow procedural task, LE QUELLENEC: improves the result of an activity a human has already completed, LE QUELLENEC: detects patterns without replacing human assessment, LE QUELLENEC: or performs a preparatory task. But there is a firm limit:
LE QUELLENEC: a system that profiles people is always high risk. NETO GALVÃO: To make that concrete for our listeners, NETO GALVÃO: a tool that extracts data from CVs into the applicant NETO GALVÃO: database is preparatory, so it is probably outside the scope. NETO GALVÃO: A tool that checks a job ad for gendered wording NETO GALVÃO: after HR has drafted it is also outside. But a tool NETO GALVÃO: that scores and ranks candidates is high risk, without a doubt. NETO GALVÃO: The same logic applies beyond hiring.
NETO GALVÃO: A tool that scores employees' performance, recommends NETO GALVÃO: who should be promoted, or flags who is likely NETO GALVÃO: to resign is profiling. It is high risk, whatever the vendor says. NETO GALVÃO: I would add one practical point. NETO GALVÃO: When a client relies on the exception, it must document NETO GALVÃO: that assessment beforehand. That is the first document the regulator, NETO GALVÃO: or the other side in litigation, will ask for. NETO GALVÃO: Another practical warning:
NETO GALVÃO: the official procurement list is rarely the full picture. NETO GALVÃO: The employer should also look for what is called shadow AI, NETO GALVÃO: meaning tools adopted directly by recruiters, NETO GALVÃO: managers, or local teams, including AI functions quietly added NETO GALVÃO: to existing HR software. The inventory should therefore cover both NETO GALVÃO: centrally approved systems and systems actually in use on the ground.
NETO GALVÃO: Now, in relation to governance, NETO GALVÃO: once the high-risk tools are identified, NETO GALVÃO: the question becomes who is responsible for what. NETO GALVÃO: What does the employer concretely have to do, Eric? LE QUELLENEC: Usually, the vendor is the provider and the employer is the deployer. LE QUELLENEC: But this can change. If the employer puts its name on the tool, LE QUELLENEC: modifies it substantially, LE QUELLENEC: or changes its intended purpose so that it becomes high risk,
LE QUELLENEC: it becomes a provider, which brings much heavier obligations. LE QUELLENEC: Before buying, the employer should check LE QUELLENEC: what the provider has done: LE QUELLENEC: the conformity assessment, CE marking, EU declaration of conformity, LE QUELLENEC: and registration in the EU database. LE QUELLENEC: That is due diligence, and it belongs in the contract. LE QUELLENEC: As a deployer, the rules are very concrete. LE QUELLENEC: Use the system in line with the instructions.
LE QUELLENEC: Assign human oversight to people with the competence, LE QUELLENEC: training, and authority to overrule it. LE QUELLENEC: Make sure the input data the employer controls are relevant LE QUELLENEC: and representative. Monitor the system, LE QUELLENEC: and if a risk appears, suspend use immediately LE QUELLENEC: and inform the provider. Serious incidents must be reported. LE QUELLENEC: Keep the logs for at least six months. LE QUELLENEC: Use the provider information
LE QUELLENEC: to carry out the GDPR data protection impact assessment (DPIA). LE QUELLENEC: Inform candidates and employees who are subject to decisions. LE QUELLENEC: Before the tool is introduced, LE QUELLENEC: inform workers' representatives and the workers concerned. LE QUELLENEC: On top of that, individuals have a right LE QUELLENEC: to an explanation of the role AI played in a decision LE QUELLENEC: that significantly affects them. A rejected candidate can ask for it. LE QUELLENEC: One common misconception concerns the fundamental
LE QUELLENEC: rights impact assessment introduced by the AI Act. It applies LE QUELLENEC: to public bodies, private entities providing public services, LE QUELLENEC: and credit and insurance uses. LE QUELLENEC: A private employer generally does not have to do it. LE QUELLENEC: For them, the key assessment remains the GDPR DPIA. NETO GALVÃO: I would add that human NETO GALVÃO: oversight must be designed into the process, not added as a label. NETO GALVÃO: The employer should define which decisions the system may support,
NETO GALVÃO: which decisions remain exclusively human, NETO GALVÃO: when a recommendation must be challenged, NETO GALVÃO: and who has the authority to stop using the tool. NETO GALVÃO: A manager who can only confirm the system's output NETO GALVÃO: is not providing meaningful oversight. In the employment context, NETO GALVÃO: implementation is also an industrial relations issue. NETO GALVÃO: Before introducing a high-risk system at work, NETO GALVÃO: the employer should identify any applicable information NETO GALVÃO: or consultation obligations toward workers' representatives.
NETO GALVÃO: The employer should explain, in accessible terms, NETO GALVÃO: what the system does, what it uses, and how it may affect workers. LE QUELLENEC: In practice, this cannot sit with one department. LE QUELLENEC: HR owns the use case; IT owns integration and security; LE QUELLENEC: the data protection officer (DPO) owns data protection and privacy; LE QUELLENEC: and legal owns the contracts. LE QUELLENEC: We recommend a small AI committee, a named owner for each tool, LE QUELLENEC: and an internal AI policy setting out authorized tools and uses.
LE QUELLENEC: The AI policy should also cover everyday generative AI. LE QUELLENEC: An HR team drafting job ads LE QUELLENEC: or appraisals with a chatbot is not high risk in itself. LE QUELLENEC: But pasting employee data into a public tool is a confidentiality LE QUELLENEC: and GDPR issue. Use approved tools only, and do not enter LE QUELLENEC: identifiable employee data unless the tool is contractually secured. LE QUELLENEC: The vendor contract is central: instructions for use,
LE QUELLENEC: information on training data and bias testing, accuracy metrics, LE QUELLENEC: access to logs, cooperation on incidents, and notice of updates. LE QUELLENEC: All of that is mandatory. LE QUELLENEC: Updates matter because of the transitional rule. LE QUELLENEC: High-risk systems in use before December 2, LE QUELLENEC: 2027, only fall under the Act LE QUELLENEC: if their design is significantly changed afterward. LE QUELLENEC: Every major update could bring a tool into scope, LE QUELLENEC: so versions must be tracked.
LE QUELLENEC: Finally, plan for switching the tool off. If it proves noncompliant LE QUELLENEC: or discriminatory, the employer must be able to stop it, LE QUELLENEC: return to a human process, and keep the logs as long LE QUELLENEC: as needed for compliance and litigation before deleting them. NETO GALVÃO: In relation to bias, bias is probably the biggest risk in HR. NETO GALVÃO: Where does it come from, and what can employers actually do about it?
LE QUELLENEC: That is not so easy to answer. It comes from several places. LE QUELLENEC: Historical bias: a tool LE QUELLENEC: trained on past hiring decisions reproduces past preferences. LE QUELLENEC: The well-known recruiting tool that penalized LE QUELLENEC: CVs mentioning women's organizations is an example. LE QUELLENEC: Representation bias: some groups are underrepresented in the data. LE QUELLENEC: Proxy bias: a postcode, a career gap, or a school name LE QUELLENEC: may be neutral on its face but stand in for origin, gender, or age.
LE QUELLENEC: Automation bias: a person validates whatever the tool suggests, LE QUELLENEC: turning human oversight into a rubber stamp. LE QUELLENEC: There's also a source people rarely think about: the question itself. LE QUELLENEC: The French Lawyers' Guide on Ethics and AI, LE QUELLENEC: adopted in March this year, draws attention to it. LE QUELLENEC: The way you phrase a question shapes the answer, LE QUELLENEC: and generative tools tend to go along with its premise. LE QUELLENEC: Ask for the best candidate for a young,
LE QUELLENEC: dynamic team, and you have introduced age discrimination LE QUELLENEC: before a single CV has been read. LE QUELLENEC: Ask whether a candidate is a good cultural fit, LE QUELLENEC: and you are asking the tool to reward similarity. LE QUELLENEC: HR prompts should be treated like a selection grid: neutral, LE QUELLENEC: written down, validated, and reviewed. NETO GALVÃO: To correct bias, you sometimes need sensitive data, NETO GALVÃO: which is where the Omnibus helps.
NETO GALVÃO: The new rules allow providers and deployers of all AI systems NETO GALVÃO: to process special categories of data to detect and correct bias, NETO GALVÃO: but only where strictly necessary NETO GALVÃO: and where synthetic or anonymized data will not do. NETO GALVÃO: That requires pseudonymization, NETO GALVÃO: security measures, and deletion once the bias is corrected. NETO GALVÃO: It is a narrow door, not an open door.
LE QUELLENEC: On top of that, employers should test outcomes, LE QUELLENEC: not just accept vendors' assurance that the tool is fair. LE QUELLENEC: That means checking whether selection, LE QUELLENEC: promotion, or performance outcomes differ materially between groups, LE QUELLENEC: investigating the reason for any disparity, LE QUELLENEC: and documenting any corrective action. LE QUELLENEC: Testing should also be repeated after significant updates LE QUELLENEC: or changes in the workforce or the use case.
LE QUELLENEC: I was walking down the streets of Paris this weekend LE QUELLENEC: and saw an advertisement LE QUELLENEC: from a famous AI provider that I want to quote: "In HR, LE QUELLENEC: a 99% accurate AI model is a 100% fail rate." NETO GALVÃO: Turning to AI literacy, NETO GALVÃO: which is the fourth step, the Omnibus softened the obligation.
NETO GALVÃO: But how far, and what does it mean for employers? LE QUELLENEC: The original text required companies LE QUELLENEC: to ensure a sufficient level of AI literacy among their staff. LE QUELLENEC: Now, they must take measures to support it. LE QUELLENEC: Lawyers will recognize the shift from an obligation of result LE QUELLENEC: to an obligation of means. NETO GALVÃO: But an obligation of means is still an obligation, NETO GALVÃO: and it follows an accountability logic. NETO GALVÃO: The employer must be able to prove it took those measures.
NETO GALVÃO: That means identifying who uses which tools, NETO GALVÃO: training them in proportion to the risk, NETO GALVÃO: and keeping records of who was trained, when, and on what. NETO GALVÃO: For high-risk tools, the bar is higher. NETO GALVÃO: The AI Act still requires the people exercising oversight NETO GALVÃO: to have the necessary competence, training, and authority. NETO GALVÃO: Put simply, a recruiter who cannot explain NETO GALVÃO: why a candidate was ranked low is not supervising anything.
LE QUELLENEC: I would add that training should be role-based. A recruiter needs LE QUELLENEC: to understand ranking errors and discriminatory proxies. LE QUELLENEC: A manager needs to know when not to rely on an AI recommendation. LE QUELLENEC: Procurement needs to ask the right vendor questions. LE QUELLENEC: Senior management needs enough information LE QUELLENEC: to allocate responsibility and resources. LE QUELLENEC: One generic awareness session will rarely be enough. NETO GALVÃO: Now, turning to recruitment in practice,
NETO GALVÃO: let's take the case clients ask about most: recruitment itself. NETO GALVÃO: How should employers frame these tools in practice? LE QUELLENEC: Everything depends on the purpose. LE QUELLENEC: If the tool stores and organizes CVs, LE QUELLENEC: there is no high-risk issue. If it helps choose between candidates, LE QUELLENEC: such as by scoring, ranking, or shortlisting, it is high risk. LE QUELLENEC: The purpose must be defined precisely, recorded, LE QUELLENEC: and respected in practice. That last part is the hard one.
LE QUELLENEC: Imagine recruiters pasting CVs into a general-purpose chatbot LE QUELLENEC: and asking it to rank them. LE QUELLENEC: The company has arguably put a general-purpose tool to a high-risk use LE QUELLENEC: and may be treated as the provider as a result. LE QUELLENEC: Hence the need to control uses and prompts. LE QUELLENEC: On the GDPR side, a DPIA will, in practice, be required. LE QUELLENEC: Evaluating and scoring candidates, LE QUELLENEC: often at scale, meets the threshold for that assessment,
LE QUELLENEC: and the French data protection authority's LE QUELLENEC: list specifically covers profiling for HR management. NETO GALVÃO: Then there is the question we always get: NETO GALVÃO: Should we ask candidates for their consent? NETO GALVÃO: Generally, no. It would be a fragile basis. NETO GALVÃO: The European Data Protection Board's (EDPB) guidelines are clear that, NETO GALVÃO: given the imbalance of power in employment, NETO GALVÃO: consent will rarely be freely given. NETO GALVÃO: A candidate who refuses may fear losing the opportunity,
NETO GALVÃO: and consent can be withdrawn at any time. NETO GALVÃO: There are better bases, such as pre-contractual measures NETO GALVÃO: or legitimate interests with a documented balancing test. NETO GALVÃO: If the decision is fully automated, NETO GALVÃO: the GDPR rules on automated decision-making will apply. NETO GALVÃO: It is allowed only if necessary for the contract, NETO GALVÃO: authorized by law, or based on explicit consent. NETO GALVÃO: The person then has the right to human intervention,
NETO GALVÃO: to express their views, and to NETO GALVÃO: contest the decision. NETO GALVÃO: The safest course is to keep a meaningful human decision at the end. NETO GALVÃO: That is why we don't NETO GALVÃO: consider consent the best lawful NETO GALVÃO: basis for these types of processing activities. NETO GALVÃO: Turning to sanctions before we close, what does a client really risk?
LE QUELLENEC: Up to 35 million euros or 7% of worldwide turnover LE QUELLENEC: for prohibited practices, such as the emotion recognition example. LE QUELLENEC: Up to 15 million euros or 3% for breaches of employer obligations, LE QUELLENEC: including transparency. For small and medium-sized enterprises, LE QUELLENEC: the lower of the two amounts applies. LE QUELLENEC: Transparency has already been in force since August of this year. LE QUELLENEC: An HR chatbot answering employees' questions must make clear
LE QUELLENEC: that it is AI, and AI-generated content must be marked. NETO GALVÃO: All of this comes on top of GDPR fines and employment litigation. NETO GALVÃO: A discriminatory tool is first and foremost a discrimination case. NETO GALVÃO: Eric, if you had to sum it up? LE QUELLENEC: Five words: map, govern, debias, train, frame.
LE QUELLENEC: Map your HR AI tools, starting from the GDPR record. LE QUELLENEC: Govern them through rules, contracts, and the ability LE QUELLENEC: to switch them off. Tackle bias, including in prompts. LE QUELLENEC: Train your people and keep proof. Frame recruitment uses precisely, LE QUELLENEC: because that is where the risks spike first. NETO GALVÃO: One last thing: the best compliance plans are built NETO GALVÃO: around the employment decision, not around the technology.
NETO GALVÃO: Start with the decision, whether hiring, NETO GALVÃO: promotion, dismissal, task allocation, or performance assessment, NETO GALVÃO: and then identify where AI influences that process. NETO GALVÃO: Make sure there is a lawful purpose, NETO GALVÃO: reliable data, meaningful human review, NETO GALVÃO: a route for the individual to challenge the outcome, NETO GALVÃO: and evidence that the process has actually worked.
NETO GALVÃO: To sum it up, the Omnibus has given employers more time, NETO GALVÃO: but it has not changed the destination. NETO GALVÃO: We have one year to do everything we have discussed NETO GALVÃO: in this podcast. Thank you very much for listening. Bye-bye. LE QUELLENEC: Thank you. Bye-bye. HOST: Thank you, Eric and Luís, and thank you all for joining us today. HOST: For our speakers' bios, upcoming podcasts, HOST: or additional information on navigating data privacy HOST: and AI in the workplace, please visit the L&E Global website.
NARRATOR: Thank you for listening to Jackson Lewis' podcast, We Get Work. NARRATOR: As a reminder, this material is provided for informational purposes only. NARRATOR: It is not intended to constitute legal advice, nor does it create NARRATOR: a client-lawyer relationship between Jackson Lewis and any recipient. NARRATOR: We Get Work is available on Apple Podcasts, Spotify, and YouTube. NARRATOR: For more information on our topics, NARRATOR: presenters, and other Jackson Lewis resources, visit jacksonlewis.com.
Federal contractors face uncertainty about whether to keep collecting race, sex, and veteran data and how to track applicants defensibly. We Get Contracting host Scott Pechaitis speaks with Pay Equity leader Laura...
Recent federal and state requirements are creating compliance pitfalls and costly exposure around worker classification, timekeeping, meal break deductions and travel time. Drawing on their experience advising...
As OFCCP's new rules and elimination of key disability data collection requirements take effect, significant affirmative action obligations remain in place. Government Contracts and Compliance Group Co-leader Scott...
The growing patchwork of AI laws makes maintaining consistent hiring practices and complying with notice, transparency and governance requirements more challenging for multistate employers. In this episode, We Get AI...
Expanded "unfair dismissal" protections that could significantly increase potential liability for multinational employers with operations in the United Kingdom are coming. In this We Get Work episode, Jackson Lewis...