The employer is responsible for algorithmic decisions
„Recruitment AI is high-risk under the EU AI Act. The employer is responsible for discrimination and the decision. The supplier's liability is contested in lawsuits. California already mandates human decision-making in dismissals.”
Counter-narrative
Some lawsuits are failing, regulations are lagging and simplifying, making liability narrow in practice. This is represented by two federal courts dismissing AI hiring discrimination cases, the Sirius XM case, a judge skeptical of Meta employees' lawsuit, and supporters of simplifying EU rules (Digital Omnibus).
Voices
EU AI Act and AI Office, California legislature, Mobley kontra Workday felperesei, EEOC and US state regulators (Connecticut, Colorado, New York), law firms, Hungarian labor lawyers
Analysis
The narrative places responsibility on the employer and attempts to involve suppliers in lawsuits. In the Workday case, the question is whether the software supplier can be held liable as the employer's agent. In the Eightfold case, the question is whether a candidate score qualifies as a consumer report. The narrative is driven by the wave of US state laws, the start of EU implementation, and class-action lawsuits. It creates a market for law firms and compliance consultants and provides a tool for the labour side. Its weak point is that regulation is patchy and slipping, and lawsuits are difficult to prove.
What's at stake: Will accountability become enforceable law or just a documentation formality?
Rival readings
- Hypothesis
- Employer liability is becoming enforceable law: in the Workday case, the court accepts that the software vendor may be liable as the employer's agent.
- Rival reading
- Lawsuits stall at the level of allegations, and courts let the supplier off. Liability remains with the employer, in practice as a documentation obligation.
- Deciding signal
- Substantive decisions in the Workday and Eightfold cases will show whether the class action proceeds and whether candidate scores qualify as consumer reports.
The split shows after 20 votes.
- Hypothesis
- The wave of US state and EU regulations is bringing real enforcement: authorities in California, Connecticut, and New York City have initiated proceedings against employers.
- Rival reading
- Regulation is slipping and simplifying: the Digital Omnibus eases or postpones the AI Act's employment obligations, leaving responsibility on paper.
- Deciding signal
- The final text of the Digital Omnibus covers deadlines for employment obligations, plus the first sanctions by the AI Office and New York City authorities against employers.
The split shows after 20 votes.
Red team
- Strongest counterargument
- Federal courts rejecting lawsuits and a judge skeptical in the Meta case state that the AI's role in decisions is hard to prove. Proponents of simplification argue that rules are narrowing, so liability is narrow in practice.
- Weakest point
- Supplier liability is supported only by litigation claims; several lawsuits failed; the AI Office's investigations are mentioned in a single professional report. The material provides no example of enforced sanctions.
- What would overturn it
- The narrative will collapse if the class-action lawsuit in the Workday case is rejected and the EU simplification delays or hollows out employment obligations.
Machine analysis based on the narrative's claim and its supporting stories. The deciding signal is an observation named in advance; later data decides between the two readings.
Analyst's note
The question is whether the employer is solely responsible for recruitment AI decisions, or if the court in the Workday case will also recognize the supplier's liability.
The full analyst's note
What it claims and what would refute it
Recruitment AI is a high-risk tool, so the employer is responsible for decisions and discrimination. This assumption is supported by EU AI Act requirements, New York regulatory clarifications, and California laws mandating human decision-making.
What to watch
One key indicator of the future direction of responsibility sharing is whether the court in the Workday case recognizes or rejects the supplier's liability.
Analytical reading
Perspective
The framework placing responsibility on employers is carried by the EU AI Office, US state regulators, and employer representatives. As part of the resistance, court decisions protecting suppliers and voices supporting regulatory simplification are emerging.
Rival reading
The story does not disclose the market-building interests of law firms and compliance consultants. The postponement of the AI Act's employment obligations to 2027 suggests that establishing enforceable liability will take longer than the current wave of lawsuits covers.
Red team
The "professional report" supporting the launch of the AI Office's investigations remains anonymous, making the actual scale of enforcement activity uncheckable. Behind the phrasing "several lawsuits failed" lie only two federal court rejections and one skeptical judicial opinion, without statistical basis.
Vakfolt
Specific risks for Hungarian employees and domestic employers receive scant attention in international narratives. Domestic coverage is minimal, with only two labor law articles addressing the issue, so local application is missing from the supporting stories.
Scenarios
- Strengthening: Employer liability strengthens if the EU AI Office launches investigations and US federal courts reject vendor protections.
- Megreked: Accountability stalls if the Digital Omnibus simplification laws enter into force and most lawsuits fail due to evidentiary difficulties.
- Megfordul: Liability shifts toward suppliers if, in the Workday case, the court treats the software manufacturer as an employer's agent and recognizes the consumer-report qualification.
Forecast log
| Claim | What counts as coming true | Deadline | Status |
|---|---|---|---|
| The employment obligations under the EU AI Act will take effect for recruitment systems in 2027. | The entry into force of obligations is confirmed by an official EU statement or implementing regulations of member states. | 2027-01-01 | nyitott |
| The California legislature passes a bill banning AI-based neural data collection and robot bosses. | The adoption of the ban law is confirmed by the California legislative journal or Governor Newsom's signature. | 2026-12-31 | nyitott |
Where it appears and how concentrated it is
21 supporting stories from 11 issues; 2 in the last four weeks. Most cited source sites: mondaq.com (6%), legal500.com (4%), hr portál (2%), hr executive (2%). The top three sites account for 12% of citations.
Machine analytical reading based on the story summary and the linked sources, following editorial rules. It contains no new facts; the facts come from the linked article. How is it made?
Narrative sub-threads
EU AI Act: high-risk recruitment
The EU AI Act classifies AI used for recruitment and selection (Annex III, point 4) as high-risk. The employer bears the burden of documentation, oversight, and transparency. The regulation becomes generally applicable from August 2, 2026, but obligations for high-risk employment systems have been delayed by approximately 15 months to 2027. An early October professional report states the EU AI Office has opened its first official investigations into recruitment and credit scoring AI systems.
EU regulation and enforcement thread: the mechanism is prior compliance and regulatory oversight, not court litigation and not national labor law.
→ lasting · evidence: strongUS lawsuits against filtering software
The Mobley v. Workday lawsuit tests whether a candidate-screening software vendor can be held liable for discrimination as the employer's agent; the court allowed a collective action. The class action against Eightfold AI disputes whether 0–5 candidate scores qualify as consumer reports under the FCRA. In October, two federal courts dismissed discrimination lawsuits over AI- and keyword-based hiring screening, and Sirius XM also secured dismissal because the plaintiff failed to show how the use of AI caused harm.
Judicial evidence and private litigation mechanism centered on supplier liability, not regulatory obligations imposed on employers.
↗ rising · evidence: strongUS state AI rules on hiring
In the absence of comprehensive federal rules, US states and cities are imposing compliance obligations on employers patchily. This includes Connecticut's AI law (effective October 1), Colorado's SB 26-189, Illinois' HB 3773, California's ADS rules, and audit requirements under New York City Local Law 144. In Colorado, the Attorney General is being urged to delay and tighten disclosure rules. According to an early November report, New York City clarified that employers are responsible for AI-assisted hiring decisions.
Member state and local legislation with audit, reporting, and entry-into-force deadlines, communicated by law firms. Not a lawsuit and not an EU regulation.
▲ at peak · evidence: strongDismissal: mandatory human decision
AI-based layoff decisions are coming under legal limits. California Governor Gavin Newsom signed SB 947 on September 30, 2026, which prohibits employers from leaving the final layoff decision solely to AI. According to one article, China ruled it unlawful to dismiss a worker because AI does their job. In the case brought by former Meta employees, the judge appeared skeptical of the evidence and pointed toward arbitration. In Hungary, the question being examined is what an employer must prove when citing AI in a redundancy.
It refers to termination of employment rather than hiring, with the mechanism being mandatory human decision and the employer's burden of proof.
↗ rising · evidence: moderateAI evaluation: transparency and review
The legal boundaries of AI-based performance evaluation and scoring are defined by the obligation to inform and human review. A survey found that 78% of managers used AI for performance evaluations, but only 16% of employees were informed about this. Spain's Royal Decree 723/2026 introduces algorithmic transparency obligations in employment relationships.
It applies to evaluation and management within existing employment relationships. The evidence is labor law analysis, member state regulations, and union demands, not recruitment or dismissal rules.
Sub-thread stories
EU AI Act: high-risk recruitment
Analyst's note
The EU AI Act classifies recruitment AI as high-risk, but employment obligations have been delayed by 15 months to 2027.
The full analyst's note
Who is affected and how
Employers bear the documentation, oversight, and transparency burdens of applying recruitment AI systems. Candidates may claim the right to human oversight, and HR departments must establish governance systems by the 2027 deadline.
What it claims and what would refute it
Under point 4 of Annex III, the EU AI Act classifies AI used for recruitment and selection as high-risk. The regulation becomes generally applicable from 2 August 2026, but obligations regarding high-risk employment systems have been delayed by 15 months to 2027.
What to watch
The 40 enforcement officers recruited by the EU AI Office start work, and the first investigation results are published before the December enforcement action.
Analytical reading
Perspective
Regulatory and professional frameworks are upheld by the European Commission, the EU AI Office, the Polish Data Protection Authority (UODO), Crowe, and the Hungarian HR trade press.
Rival reading
Simplification debates following the Digital Omnibus suggest that rule amendments may be justified not only by risk management but also by their impact on worker safety. According to supplier comparisons, due to market competition mechanisms, liability may remain with the employer rather than regulatory authorities.
Red team
The October 5, 2026 professional report mentions the first investigations by the EU AI Office, while the sources cite the hiring of 40 enforcement officers before the December action.
Vakfolt
The story mentions the postponement of rules affecting Irish employees. Hungarian-language summaries do not detail how the regulation affects Hungarian subsidiaries of non-EU headquartered companies.
Scenarios
- Strengthening: The EU AI Office would launch further investigations into recruitment systems if more complaints are received before the December action.
- Megreked: Compliance enforcement will slow down if member states do not adopt the checklists before the 2027 deadline.
- Megfordul: Recruitment AI systems would be classified into a lower risk category if the simplification of rules is adopted following the Digital Omnibus.
Forecast log
| Claim | What counts as coming true | Deadline | Status |
|---|---|---|---|
| The EU AI Office publishes its first investigative report on recruitment AI systems. | The report is publicly available on the official website of the EU AI Office or in a professional report cited by the Hungarian HR trade press. | 2027-03-15 | nyitott |
| The Polish data protection authority (UODO) is issuing a new checklist for HR systems. | The official document will appear on the website of the Polish Data Protection Authority or in the comparison published by Crowe. | 2027-01-30 | nyitott |
Where it appears and how concentrated it is
11 supporting stories from 11 issues; 1 in the last four weeks. Most cited source sites: hr executive (5%), zendata.it (2%), hr portál (2%), rsmanagement.ai (2%). The top three sites account for 9% of citations.
Machine analytical reading based on the story summary and the linked sources, following editorial rules. It contains no new facts; the facts come from the linked article. How is it made?
US lawsuits against filtering software
Analyst's note
Software vendors' liability depends on whether courts in Mobley v. Workday recognize legal risk for vendors acting as agents.
The full analyst's note
Who is affected and how
Suppliers of candidate screening software, such as Workday and Eightfold AI, may face direct legal risks due to liability as employers' agents. Both candidates and employers are affected, as courts in some cases allow class-action lawsuits but reject them in others if discrimination cannot be proven.
What it claims and what would refute it
The Mobley v. Workday case examines whether a software vendor can be held liable for discrimination when acting as an employer's agent. The lawsuit against Eightfold AI claims that candidate scores between 0 and 5 qualify as consumer reports under the FCRA; the company has requested dismissal.
What to watch
The most significant observable signal is the court's ruling on supplier liability in the Mobley vs. Workday case.
Analytical reading
Perspective
Plaintiffs and legal experts focus on algorithmic bias in AI tools and supplier liability within private litigation mechanisms. Sources report facts from courts regarding rejected or approved lawsuits.
Rival reading
The dismissal of the Sirius XM case and others shows that litigation risk stems not from the use of AI technology but from the lack of proof of specific discrimination. The interests of employers and software vendors may differ, as vendors may point to shifting liability, while employers may cite the lack of regulation as a defense.
Red team
The weakest point of the evidence is the causal link between AI use and specific harm in the Sirius XM case and others. The sources record legal claims and the outcomes of court decisions.
Vakfolt
Sources do not address how software vendors protect against liability shifting in contracts.
Scenarios
- Strengthening: The number of class actions will rise if courts recognize vendor liability in Mobley v. Workday.
- Megreked: The number of lawsuits will stagnate if courts continue to reject claims that fail to demonstrate specific harm.
- Megfordul: Legal risk will fall if federal regulation creates a comprehensive framework in place of private-litigation mechanisms.
Forecast log
| Claim | What counts as coming true | Deadline | Status |
|---|---|---|---|
| In the Mobley v. Workday case, the court will either reject or accept the class action. | The court ruling is published on legal500.com or reuters.com. | 2027-03-15 | nyitott |
| In the lawsuit against Eightfold AI, the court will decide whether the scores qualify as consumer reports. | The court ruling or the settlement between the parties will be published on hrdive.com. | 2027-02-28 | nyitott |
Where it appears and how concentrated it is
9 supporting stories from 9 issues; 1 in the last four weeks. Most cited source sites: legal500.com (9%), mondaq.com (7%), hrdive.com (4%), reuters.com (4%). The top three sites account for 20% of citations.
Machine analytical reading based on the story summary and the linked sources, following editorial rules. It contains no new facts; the facts come from the linked article. How is it made?
US state AI rules on hiring
Analyst's note
In the absence of federal rules, US states and cities are introducing their own audit and notification obligations for employers, with differing deadlines.
The full analyst's note
Who is affected and how
State and city regulations impose compliance obligations patchily on US employers, including Connecticut's AI law, Colorado's SB 26-189, Illinois' HB 3773, California's ADS rules, and the audit requirement under New York City Local Law 144. Audit, disclosure, and effective date deadlines for state and local legislation are communicated by law firms.
What it claims and what would refute it
In the absence of comprehensive federal rules, US states and cities are introducing their own audit and reporting obligations for employers, with different deadlines and requirements. This assumption would be refuted if the federal government adopted a unified law or if local authorities revoked previously introduced regulations.
What to watch
The number of legal disputes and employer complaints filed in the first month following the October 1 effective date of the Connecticut AI law.
Analytical reading
Perspective
The narrative is carried by law firms and legal experts analyzing the legislatures of Connecticut and Colorado, as well as New York City's measures. The sources focus on employer and vendor liability, as well as compliance obligations.
Rival reading
Regulatory diversity may be a response to local political pressure in addition to reducing legal risk: legislators may prioritize voter satisfaction over technological progress. Behind this may lie the fact that states compete for investors and technology companies, so the strictness of rules may align with the economic competitive situation.
Red team
The source material is dominated by legal summaries and analyses from law firms, whose business interests may bias the emphasis toward the necessity of compliance obligations. The sources contain no independent, empirical data on the actual effectiveness of regulations or the reduction of discrimination.
Vakfolt
The supporting stories exclude small businesses and resource-constrained employers, for whom compliance costs across multiple states could be disproportionately high. The social consequences of regulation, such as declining employee trust or slower recruitment processes, are not discussed.
Scenarios
- Strengthening: More states will adopt similar AI recruitment laws, and audit requirements under existing rules will tighten, as indicated by the spread of legal literature.
- Megreked: The regulatory wave will slow down if the federal government intervenes, or if employers resist further compliance costs due to the economic environment.
- Megfordul: Local authorities review and relax previously introduced regulations if they adversely affect the local labour market or businesses.
Forecast log
| Claim | What counts as coming true | Deadline | Status |
|---|---|---|---|
| In the first quarter after the Connecticut AI Act entered into force, at least five legal cases are initiated against employers for violations of the rules. | The number of legal decisions and complaint filings published by local courts or state legal protection offices. | 2027-01-31 | nyitott |
| The Colorado Attorney General delays or modifies AI hiring notification rules despite the November urging. | An official statement from the Colorado Attorney General's office or an amendment adopted by the legislature. | 2026-12-31 | nyitott |
Where it appears and how concentrated it is
8 supporting stories from 8 issues; 3 in the last four weeks. Most cited source sites: knowledge.dlapiper.com (4%), njworkplacelawblog.com (4%), natlawreview.com (4%), mvalaw.com (4%). The top three sites account for 11% of citations.
Machine analytical reading based on the story summary and the linked sources, following editorial rules. It contains no new facts; the facts come from the linked article. How is it made?
Dismissal: mandatory human decision
Analyst's note
California's SB 947 law prohibits employers from relying solely on AI for final termination decisions; this could be overturned by a court ruling or the legislature.
The full analyst's note
Who is affected and how
California employers must insert human decision-making into the final stage of the dismissal process. The ban on purely algorithmic decisions strengthens legal protection for workers, and the regulation fulfills union demands.
What it claims and what would refute it
California's SB 947 law prohibits employers from relying solely on artificial intelligence for final termination decisions. This claim would be refuted if future court rulings invalidate the law or if lawmakers lift the restrictions for economic efficiency.
What to watch
Initial rulings by California courts on the application of SB 947, particularly the outcome of the Meta workers' case in arbitration proceedings.
Analytical reading
Perspective
Within a framework built around worker protection and the necessity of human decision-making, the voices of the California legislature and trade unions dominate. In international comparison, Hungarian labor law experts and the Chinese regulatory approach also appear.
Rival reading
The main driver of layoffs may be the maximization of economic efficiency, where AI serves as a cost-cutting tool and is not the sole reason for the decision. Regulation may increase the administrative burden on employers alongside protecting workers, at the expense of competitiveness.
Red team
In the lawsuit involving Meta employees, the judge skeptically received the evidence presented, which questions the direct causal link between AI and layoffs. The judge dismissed the case during the arbitration proceedings.
Vakfolt
The impact assessment of the regulation does not include small businesses and employers in non-technology sectors, even though they may also use AI for workforce reductions.
Scenarios
- Strengthening: The regulation will extend to other states or the international level if the California model reduces legal disputes and increases employee satisfaction.
- Megreked: Implementation of the law will slow down if companies face legal challenges or technical difficulties in practically introducing human decision-making.
- Megfordul: The scope of the regulation would be reduced or lifted if, during an economic downturn, employers pressure lawmakers to allow more flexible workforce management.
Forecast log
| Claim | What counts as coming true | Deadline | Status |
|---|---|---|---|
| A California court rejects the lawsuit by Meta workers, citing the arbitration route. | The court judgment or arbitration award will appear in the official court register or in a press release. | 2027-03-15 | nyitott |
| Another US state introduces a restriction on AI-based dismissals similar to California's. | Adoption of the new bill or the governor's signature appears in the given state's official statute book. | 2027-05-01 | nyitott |
Where it appears and how concentrated it is
10 supporting stories from 8 issues; 1 in the last four weeks. Most cited source sites: news.bloomberglaw.com (7%), engadget.com (7%), smartbrief.com (7%), carltonfields.com (4%). The top three sites account for 21% of citations.
Machine analytical reading based on the story summary and the linked sources, following editorial rules. It contains no new facts; the facts come from the linked article. How is it made?
AI evaluation: transparency and review
Analyst's note
According to Spanish Regulation 723/2026 and trade unions, the limit of AI-based assessment is worker notification and human review.
The full analyst's note
Who is affected and how
Workers are affected by decisions made by performance-evaluation algorithms, which often remain hidden from them. Managers and employers are interested in the efficiency of the tools, while trade unions and legal experts demand the right to information and human review. Spanish government decree 723/2026 imposes new obligations on employers, while Hungarian legal literature deals with setting the boundaries.
What it claims and what would refute it
The legal boundaries of AI-based assessment are defined by the obligation to inform and the necessity of human review; Spain's Royal Decree 723/2026 and union demands cite this. The assumption would be refuted if evidence emerged that employers can lawfully enforce algorithmic decisions that require neither human intervention nor informing workers.
What to watch
The publication of Hungarian draft laws or the effective date of the Spanish regulation may signal changes in the regulatory environment.
Analytical reading
Perspective
The narrative is carried by Hungarian labor law experts, the Spanish government, and trade unions, focusing on workers' legal protection and transparency of control within employment relationships. The voice of the Hungarian HR press and domestic employers is missing, so there is little information on the difficulties of practical implementation and companies' internal interests.
Rival reading
AI usage by 78% of leaders may be explained by technological progress and market competition pressure, with efficiency gains and cost reduction as the primary drivers. Employers may use these systems for faster and more objective decision-making, even if communication is lacking.
Red team
The source of the survey, which shows 78% of leaders and 16% of informed employees, does not detail the representativeness of the sample or the questionnaire methodology. The concept of the "invisible boss" and the description of Co-op customer service cases are based on subjective observation, which is difficult to verify with independent data.
Vakfolt
Supporting stories exclude small businesses and non-unionized workers, who are less directly affected by the Spanish regulation and Hungarian legal disputes. In the focus on conflict between employers and employees, the interests and risks of technology providers and algorithm developers remain absent.
Scenarios
- Strengthening: Pressure from Hungarian legal literature and unions leads the government to introduce similar transparency obligations, signaled by the publication of draft legislation.
- Megreked: Employers continue not to disclose AI usage because legal consequences are absent; this is indicated by the lack of reporting obligations and missed reviews.
- Megfordul: Technological progress renders human oversight obsolete, as indicated by the recognition of the legal status of algorithms and the cessation of informing employees.
Forecast log
| Claim | What counts as coming true | Deadline | Status |
|---|---|---|---|
| The Hungarian government announces a new draft law on algorithmic transparency regarding employment relationships. | The draft law will officially appear on the government portal or in the Hungarian Gazette (Magyar Közlöny). | 2027-03-15 | nyitott |
| The first fines or legal decisions against employers under Spain's 723/2026 regulation will appear. | The fine or court ruling is published in the Spanish legal register or in the press. | 2027-01-30 | nyitott |
Where it appears and how concentrated it is
7 supporting stories from 5 issues; 1 in the last four weeks. Most cited source sites: arsboni.hu (8%), hirnavigator.hu (8%), fulophenrik.hu (4%), startlap.hu (4%). The top three sites account for 21% of citations.
Machine analytical reading based on the story summary and the linked sources, following editorial rules. It contains no new facts; the facts come from the linked article. How is it made?
Google search interest
Hungary
| Topic and search term | Weekly | Average |
|---|---|---|
| The employer is responsible for algorithmic decisionsalgorithmic discrimination | 0 | |
| EU AI Act: high-risk recruitmenteu ai act | 4.1 | |
| US lawsuits against filtering softwareworkday per | 0 | |
| US state AI rules on hiringAI regulation | 0 | |
| Dismissal: mandatory human decisionunlawful dismissal | 0 | |
| AI evaluation: transparency and reviewperformance evaluation | 0 |
World
| Topic and search term | Weekly | Average |
|---|---|---|
| The employer is responsible for algorithmic decisionsalgorithmic bias | 1.8 | |
| EU AI Act: high-risk recruitmenteu ai act | 20.8 | |
| US lawsuits against filtering softwaremobley v workday | 0 | |
| US state AI rules on hiringai hiring laws | 0 | |
| Dismissal: mandatory human decisionwrongful termination | 11.5 | |
| AI evaluation: transparency and reviewperformance review | 623.2 |
Google Trends relative search index over the period studied, weekly average. A narrative and its sub-threads appear in a single comparison: 100 is the group's most-searched day, and the figures can only be compared within the group. The search term was chosen by machine analysis; it measures how much the term is searched, not how widespread the narrative is.
Who is affected and how
- HR professional
- Employers are responsible for AI-supported hiring decisions. The EEOC applies the adverse impact doctrine to AI tools, meaning unintentional discrimination can also create liability. It is advisable to inventory all AI used in HR and document bias audits and human oversight.
- manager
- In California, SB 947 already prohibits AI from making final termination decisions alone. It is advisable to document who approves the decision and on what grounds.
- SME owner
- Using a vendor tool does not absolve the employer of liability; it is advisable to request audit rights and anti-discrimination liability clauses in the contract.
- job seeker
- The right to information is expanding, for example under the AI Act's transparency obligations and a Spanish decree. In California, human decision-making is mandatory during layoffs. It is advisable to ask after a rejection whether AI was used and request human review.
Hungary
Domestic coverage is minimal. Two Hungarian labor law articles address the topic. Henrik Fülöp's August 21, 2026 article discusses how far AI may decide autonomously in performance evaluations, when human review is required and what information must be provided. An August 6 article reviews what an employer must prove when citing AI in layoffs.
Weekly trend
Turning points
Related narratives in the News magazine
Supporting stories
Machine analysis based on story titles and summaries, with a verification pass. The source of the facts is the article linked at each story.
What do you make of this piece?
The split shows after 20 votes.
Why? Your answer is not shown publicly; it helps the editors.