The following article first appeared on the Employment Law Letter section of Shipman & Goodwin’s website. Reprinted here with permission.
If your organization uses AI or automated tools to screen, rank, or evaluate job candidates, it’s worth paying attention to recent developments in the Northern District of California and the Connecticut State Legislature.
California’s Mobley v. Workday, Inc., No. 23-CV-00770-RFL, now in its third year of litigation, continues to issue important decisions that address the question of how existing federal antidiscrimination laws apply to AI-driven employment systems.
In Connecticut, the state Legislature recently passed SB 5, the Connecticut Artificial Intelligence Accountability and Transparency Act, which reflects a growing trend toward greater regulation of the use of automated employment recruiting tools.
Taken together, these developments confirm that the legal risks associated with AI in employment are real, growing, and no longer theoretical.
Mobley vs. Workday Overview
At the heart of this case is that it raises direct questions that should concern any employer who uses automated recruiting tools. Who is responsible if an AI-powered selection system produces discriminatory results?
Workday, Inc. provides a platform used by thousands of employers to process job applications and often serves as the gateway through which applicants must pass to be considered for employment.
Plaintiffs allege that the platform’s AI-driven screening and ranking tools produce discriminatory results.
The important facts are:
- The lawsuit was originally filed in 2023 by Derek Mobley, an African-American applicant over the age of 40, who claims he was repeatedly denied a job through Workday’s platform. The complaint has since been expanded to include additional plaintiffs across multiple protected classes.
- This claim invokes Title VII of the Civil Rights Act, the Age Discrimination in Employment Act, and the Americans with Disabilities Amendments Act, as well as numerous state law assertions. The underlying theory is algorithmic bias, meaning that AI systems can produce discriminatory outcomes even when no one intends to discriminate.
- The suit says Workday’s AI systems are trained on past hiring data that can reproduce patterns embedded in past decisions. Even if a protected characteristic is not used directly as input, the system may rely on surrogate data. Employment gaps, years of experience, education, and career trajectories can all be correlated with characteristics such as disability, age, race, and gender. The complaint further alleges that Workday’s integration of HiredScore’s AI recruiting tools in 2024 will further incorporate automated scoring into the decision-making process.
- Plaintiffs also took issue with the personality and assessment tools integrated into the platform, arguing that these are data collection instruments that feed broader candidate profiles rather than neutral screening mechanisms. The complaint alleges that in some cases, these assessments can act like medical questions prohibited by the ADA because they can infer disability through proxy indicators such as illness-related absenteeism or health-related behavior patterns.
Current status of litigation
The case is now more than three years old, and the court has issued several important decisions.
In 2024, the court rejected the intentional discrimination claim but allowed the disparate impact claim to proceed. This is a distinction with significant practical implications (more on this below).
The complaint has since been amended multiple times. Most recently, the plaintiffs filed a Third Amended Complaint on March 27, 2026, adding new named plaintiffs, expanding race claims to include Asian Americans, expanding disability ratings to include physical disabilities, and strengthening California Fair Employment and Housing Act claims with specific allegations that Workday’s screening system was designed, trained, and administered by its California headquarters.
Workday filed its latest motion to dismiss on April 10, 2026, arguing that the plaintiffs exceeded the limits of the court’s leave to amend, that certain claims were time-barred, and that the FEHA claims remained insufficient. The motion has now been fully explained and awaits a decision.
Regardless of how this motion is resolved, the key points for employers are: The core claim about federal disparate influence is that it has survived all odds.
The core federal disparate influence argument has survived all challenges to date.
Workday is also seeking interlocutory certification on the issue of whether job applicants can make disparate impact claims under the ADEA. This is an indication that the legal issues in this case are being contested at the highest level and remain unresolved.
This distinction is important. Disparate impact claims focus on outcomes rather than intentions.
If you unfairly single out members of a protected group during the hiring process, you may be subject to liability even in the absence of evidence of discriminatory motive, unless you can prove that the conduct was job-related and consistent with business necessity.
Plaintiffs also point to audit data that purports to show statistically significant differences affecting certain groups.
If that evidence is valid, it could play an important role as the case progresses, particularly as it relates to class and class certification.
Why this matters to your organization in Connecticut
Mobley’s case is pending in California, but his legal problems are spreading across the country.
Additional risks exist, especially for Connecticut employers. Connecticut’s Fair Employment Practices Act provides broad protections against employment discrimination, and in some cases, it is sufficient for a plaintiff to prove that the discrimination was a “motivating factor” and a lower standard than required by federal law.
In addition, on May 1, 2026, the Connecticut General Assembly passed the Connecticut Artificial Intelligence Accountability and Transparency Act.
Organizations can still be held liable if they use an AI recruitment tool and it produces discriminatory results.
Among other things, SB 5 would require employers who use AI tools that significantly influence hiring decisions to provide plain language disclosures and pre-decision notice to applicants and employees before those tools are used to make decisions.
The bill also amends CFEPA to clarify that reliance on automated tools is not a defense to discrimination claims.
We published a detailed breakdown of SB 5’s employer provisions, effective date, and compliance steps in a separate post.
Simply put, an organization can still be held liable if an AI recruiting tool is used to produce a discriminatory outcome, even if the tool was developed by a reputable vendor, had no intent to discriminate, and even if it did not know how the tool worked.
What Connecticut Employers Should Do Now
The lesson from Mobley and Connecticut’s new AI bill is not to abandon AI in employment, but rather to use it with proper oversight, even though these tools can bring real efficiencies. Here are some specific steps to consider:
- Know the tools. If AI is involved in screening or ranking candidates, you need to know what data is being used and how decisions are being made. In black-box systems, the logic behind decisions is opaque even to employers, putting them at significant risk.
- Review your relationships with vendors. Mobley’s story illustrates how quickly AI recruiting platforms are evolving, as Workday’s 2024 HiredScore integration adds an additional layer of AI-driven processing to vet. Review your contract to understand what representations the vendor has made regarding the compliance of their tools, what testing has been performed, and who will be held accountable if issues surface. Indemnification clauses, audit rights, and data sharing obligations all deserve close attention.
- Audit your evaluation. If your hiring process includes evaluations or personality tests, consider whether any of them could be characterized as disability-related inquiries under the ADA. The line between permissible pre-employment screening tools and impermissible medical investigations is one that courts have actively defined.
- Be prepared to justify the process. If the tool has a disparate impact, you need to prove that it is job-relevant and consistent with business need. This analysis should be done up front, rather than only after a claim is filed.
- Document your diligence. Document the steps you take to evaluate and monitor your AI recruitment tools, including vendor interactions, audit results, and bias testing. If something goes wrong with the process, documentation is the best evidence of good faith.
- Prepare for SB 5 compliance. If your organization uses AI tools for recruiting and other employment decisions, start mapping those tools today. Identify systems that produce output that has a significant impact on decision-making, review vendor agreements, and begin building disclosure and notification workflows. A separate post on SB 5 will provide a detailed compliance roadmap.
Mobley v. Workday is one of the first major cases to question whether existing anti-discrimination laws can hold AI recruiting tools accountable.
Connecticut’s SB 5 represents the legislative side of the same trend.
Employers who take the time now to understand their tools, audit them for bias, and build compliance workflows will be in a much stronger position when these issues come to them, or when they arrive.
About the author: Justin Cedeño is an associate in Shipman’s Employment, Labor and School Law practice group.
