The Workday Lawsuit That Shocked the World

A recent class action lawsuit in the United States has stirred global attention: plaintiffs allege that a recruiting algorithm used by Workday discriminated against Black, over-40, and disabled applicants. The case has made international headlines — and for good reason.

While the lawsuit currently targets the software developer, it’s crucial to understand that AI is not a closed box — and responsibility doesn’t stop with the vendor.

The EU AI Act clearly states: the entire supply chain is accountable. If an organization uses an AI system that plays a role in employment decisions, that system is classified as high-risk.

In such cases, the AI deployer — i.e. the employer — is legally obligated to:

  • Review the technical documentation
  • Evaluate the system’s purpose and context
  • Identify and mitigate risks through impact assessments
  • Measure fairness and non-discrimination
  • Ensure human oversight is in place

These are not just best practices — they are legal obligations under the EU AI Act. Moreover, AI doesn’t operate in a vacuum. In Hungary, for example, both the Equal Treatment Act (2003. évi CXXV.) and the Labour Code prohibit discrimination, regardless of whether it’s committed by a human or an algorithm.

Key takeaway: this lawsuit could easily target any Workday customer — even today, and even within the EU.

Where can AI go wrong?

It’s unlikely that Workday set out to create a discriminatory system — yet it still happened. How?

  • Non-representative training data: if historical data is biased, the model will be too.
  • Flawed target variables: if the goal (e.g. “who was hired”) reflects biased past decisions.
  • Hidden proxy variables: the model may not use gender, age, or disability explicitly but may infer them through proxies (e.g. school, location, vocabulary).
  • Lack of oversight: without regular audits or human review, bias may go unnoticed.

I explore these issues in more detail in my earlier article “The Dark Side of AI: Hidden Biases in Algorithms”.

Diversity as a cornerstone of AI governance

Creating safe and fair AI systems requires diverse teams:

  • Interdisciplinary (HR, IT, Legal, Data Privacy, etc.)
  • Socially diverse
  • AI-literate — able to understand how AI works, where its limits lie, and what risks it entails

But what about women in the AI age?

According to the International Labour Organization (ILO), generative AI is most likely to automate administrative and assistant roles — positions where women are disproportionately represented. This means gender equality achievements are at risk (Gmyrek et al., 2025).

That’s why career transitions, reskilling strategies, and access to learning opportunities must be designed consciously and equitably.

We must ask:

  • Who is responsible for reskilling?
  • Is learning time offered during working hours?
  • Can those with caregiving responsibilities (typically women) realistically participate?

And if we’re the ones developing AI, it’s even more critical that our teams are diverse. Sadly, we’re far from that goal: only 22% of AI engineers are women, according to UNESCO, and representation of marginalized groups is similarly low (Gmyrek et al., 2025).

Final thought: to preserve DEIB, we must practice it in AI governance

This isn’t just a catchy phrase — it’s the key to the future. AI is not just technology; it reflects and amplifies our human decisions.

If we want AI to reduce, rather than reinforce, inequalities, we must:

  • Begin mapping AI’s impact now and design reskilling strategies with DEIB principles embedded
  • Invest in AI literacy — not just as a best practice, but as a legal requirement under the EU AI Act
  • Establish diverse AI governance structures and development teams

If DEIB is truly a goal, then it’s time to move beyond empty slogans. In the age of AI, embedding DEIB into AI governance is the real starting point.

Deploying hiring AI? The accountability is yours — evidence it.

Independent bias audit →AI Governance Quick Scan →