AI screening in hiring is no longer just a procurement question. It is now a litigation and compliance question. In one week in November 2026, Workday asked a federal court to break up the largest AI hiring bias case in the US. Law firms also published guidance on Connecticut’s 2027 employment-AI rules and California’s new limits on “robo bosses”. The real risk sits with how each employer configures and supervises its tools, not with the vendor’s marketing claims. HR leaders should spend the next 90 days building an inventory, a paper trail and a human-review step for every automated decision.
Two years ago, the main question about AI in recruiting was whether the tools were any good. In late 2026 the question is who answers for them when they go wrong. The week of November 10 made that plain. In Mobley v. Workday, the vendor filed its opposition to class certification. Around the same time, a run of law-firm alerts covered new state rules that take effect in 2027. This article draws on news coverage, legal commentary and search-interest data to explain what changed, why it matters and what an HR team can do about it before the next budget cycle.
For the filing-level detail on the Workday case, see our earlier briefing: Workday Opposes Class Certification in AI Hiring Bias Suit: Notes for HR. This piece covers the bigger picture.
This is not legal advice. Consult employment counsel on your specific obligations.
What the headlines are saying
We reviewed about 30 English-language news and legal-commentary items from the last two weeks, pulled from Google News RSS feeds on AI hiring bias, AI hiring regulation and the Workday case. Three themes dominate.
- The vendor-liability fight. Reuters, Law360, Forbes, MLex and Seeking Alpha all led with Workday’s argument. In Workday’s telling, its customers make the hiring decisions, the challenged features work differently from one employer to the next, and millions of applicants cannot sue as a single group. The hearing on class certification is set for March 9, 2027.
- Mixed early results for plaintiffs. HRMorning and HR Dive reported that Sirius XM won the first round of Harper v. Sirius XM on September 30, 2026. The court found that the plaintiff had not shown how the employer’s alleged use of AI discriminated against him. Commentators read the ruling as a reprieve, not an acquittal. It suggests that in these cases, both plaintiffs and defendants will depend on detailed records of each applicant’s path through the process.
- A growing patchwork of state rules. Foley & Lardner, Cooley, JD Supra and the Connecticut Business & Industry Association covered Connecticut’s new employment-AI obligations. Ogletree and Davis Wright Tremaine covered California’s 2027 employment laws, including limits on relying only on automated systems for discipline and termination. Commentary on New York City’s Local Law 144, the bias-audit rule for automated employment decision tools in force since 2023, keeps returning to one point: audits that exist only on paper do not protect an employer.
Two other threads sit in the background. The New York Times and The Washington Post both framed the November 2026 midterms as a turning point in public attitudes toward AI, with voters rejecting AI-related measures, including data-center projects, in several places. The Guardian reported that public-sector unions are winning contract language that limits how AI can replace or monitor workers. Neither story is about recruiting directly. Both point to the same climate: less public patience with automated decisions about people.
What Google Trends shows
We pulled 90 days of worldwide Google Trends data (August 13 to November 13, 2026) for “AI hiring”, “AI hiring bias” and “Workday lawsuit”. The picture is useful mainly because it is so quiet.
Three readings stand out:
- The public is not following the court case. “Workday lawsuit” barely moved during the week of the class-certification filing, sitting at an index of 1 to 2. Lawyers, vendors and HR buyers care about this case. Ordinary job seekers mostly do not. Don’t expect public pressure to drive change here. The pressure will come from courts and regulators.
- People search for tools, not risks. The top related queries for “AI hiring” were “ai in hiring”, “ai hiring software”, “ai hiring platform” and “ai hiring tools”. Buying interest is running well ahead of interest in bias or compliance. That gap is where governance problems start.
- Interest is concentrated in Asia and the UK. The Philippines, South Korea, Singapore, Bangladesh, Kenya, Hong Kong and the UK all indexed higher than the US for “AI hiring”. These are big outsourcing and shared-services markets. Multinationals that screen candidates there should expect US-style litigation habits to show up eventually, and EU and UK data-protection rules already apply.
One caveat: every series we pulled showed a common spike around September 20 to 21, 2026. That looks like a sampling artifact in the data, so we compare four-week averages rather than single-day peaks.
Why the Workday argument matters even if Workday wins
Workday’s defense is simple to state. Its customers decide whom to hire. Each customer configures the screening features differently. Three challenged features were built by different teams on different data. So there is no single “common question” that would let thousands of applicants sue together.
That may be a strong litigation position. It is also a warning to every employer. The more a vendor argues that the outcome depends on the customer’s configuration, the more the outcome depends on the customer. If courts accept that framing, the next wave of claims could target employers directly, one implementation at a time. Each employer would then need to show what thresholds it set, which signals it weighted, who reviewed the shortlist and why a given candidate was rejected.
The Harper ruling points the same way. The claim failed at an early stage because it lacked specifics about how the tool caused the harm. Plaintiffs’ lawyers will learn from that. Future complaints are likely to arrive with more detail about the tools and better statistics. Some of that information will come from discovery in cases like Mobley, which has already run for 18 months.
The 2027 compliance map, in plain terms
The details differ by state, and some reports conflict on effective dates and bill names, so check the enacted text with counsel. The direction of travel is consistent, though.
| Jurisdiction | What it asks of employers (as reported) | Timing |
|---|---|---|
| Connecticut | Plain-language disclosure when applicants or employees interact with automated employment decision technology. Notice when AI makes or materially influences decisions. An emphasis on explanation, correction and human review over mandatory bias testing. WARN notices must say whether layoffs are linked to AI. | Main duties from October 1, 2027. WARN disclosure reported from October 1, 2026. |
| California | Successor to the vetoed 2025 “No Robo Bosses” bill. Limits relying only on automated decision systems for discipline, termination or deactivation, with notice obligations. Separate CPPA rules on automated decision-making and Civil Rights Council rules also apply. | Reported effective date in 2027. Verify against the enacted text. |
| New York City | Annual independent bias audits for automated employment decision tools, published summaries and candidate notice. | In force. Amendments are under discussion. |
| Federal (US) | No AI-specific hiring statute. Title VII, the ADEA and the ADA still apply to discriminatory outcomes, however a decision was made. | Ongoing. |
Taken together, these rules ask for four things: know what tools you use, tell people, keep a human in the loop and keep records. None of that requires deep technical expertise. All of it requires coordination across HR, legal, IT and procurement, which is why it so often slips.
A 90-day plan for HR leaders
Days 1–30: inventory and ownership
- List every tool that scores, ranks, filters, schedules or recommends candidates or employees. Include features inside your ATS or HRIS that were switched on by default.
- For each tool, record who configured it, when, and what knockout questions, thresholds or weightings are active.
- Name one accountable owner for each tool. A committee is not an owner.
Days 31–60: contracts and evidence
- Review vendor contracts for indemnification, limitation-of-liability, audit rights and data-retention clauses. If the vendor’s defense is “the customer decides”, the contract should at least guarantee you access to the logs you would need to defend yourself.
- Run or commission a disparate-impact analysis on recent requisitions, broken down by job family, and keep the methodology on file.
- Check whether your candidate notices match how the tools actually work, not how the sales deck describes them.
Days 61–90: human review that holds up
- Define which decisions always need a person: final rejection after interview, any termination or discipline, and any automated knockout in high-volume roles.
- Make the review real. A reviewer who approves 400 AI shortlists an hour is not providing meaningful oversight, and regulators increasingly know it.
- Build a simple route for candidates to ask for review or correction. Connecticut’s approach centers on this right, and it is cheap to offer everywhere.
The global angle
Most coverage of AI hiring rules is American, but kurums.com readers hire across borders. The EU AI Act treats AI systems used for recruitment and worker management as high-risk, with obligations phasing in over 2026 and beyond. The UK has no AI-specific employment statute, but UK GDPR rules on solely automated decisions apply. Australia’s government announced plans in November 2026 for mandatory safety rules on AI, according to The Conversation. Search interest shows employers in Asian hubs are actively shopping for AI hiring software. Some of those markets have weaker local rules, but their candidates often end up in pipelines run by multinationals that answer to US or EU regulators.
The practical answer is to build to the strictest standard you face and then apply it everywhere. Running one notice, one review process and one audit method is cheaper than maintaining five versions.
What to watch next
- December 8, 2026: Plaintiffs’ reply briefs are due in Mobley v. Workday.
- March 9, 2027: The class-certification hearing before Judge Rita F. Lin in the Northern District of California.
- Through 2027: Connecticut implementation guidance and California’s final rule text, plus any changes to NYC Local Law 144.
- Vendor behavior: Watch for HR-tech vendors rewriting customer contracts to push configuration risk onto employers, or competing on audit-ready logging.
Frequently asked questions
Has any court found an AI hiring vendor liable?
Not in the cases discussed here. Mobley v. Workday is still at the class-certification stage, and Harper v. Sirius XM was dismissed at an early stage. The rulings so far are about procedure and how much detail plaintiffs must plead, not about whether the tools discriminate.
Should we stop using AI screening?
Not necessarily. Unstructured human screening has well-documented bias problems too. The goal is screening you can explain, monitor and defend, whoever or whatever does it.
Sources and further reading
- Reuters, Law360, Forbes, MLex and Seeking Alpha coverage of the Mobley v. Workday class-certification opposition, November 10–12, 2026.
- HRMorning: Sirius XM Wins Dismissal of AI Hiring Bias Lawsuit and HR Dive.
- Foley & Lardner: Connecticut’s 2027 Rules for AI in Employment; Cooley: Connecticut’s AI Rules Take Shape.
- Ogletree and Davis Wright Tremaine alerts on California’s 2027 employment laws; National Law Review on the No Robo Bosses Act.
- The Guardian: “US public workers are fighting to protect their jobs from AI – and some are winning”, November 12, 2026.
- Google Trends: “AI hiring”, past 90 days.
Discover more from Kurums | Business Intelligence
Subscribe to get the latest posts sent to your email.