Workday faces a lawsuit claiming its hiring AI discriminates against Black applicants by down-weighting their resumes

A group of job applicants sued Workday. They allege that Workdays AI tools helped screen resumes for open positions. The tools automatically ranked candidates based on their resumes. The lawsuit claims the AI penalized Black applicants. It likely did this by down-weighting their resumes or filtering them out. The mechanism is the model learning historical bias from past hiring data. The AI likely noticed that certain groups were underrepresented in past hires and decided to lower their scores.

This is an algorithmic discrimination failure. The AI showed disparate impact. It made adverse decisions without human review. It lacked explainability. The company did not catch the bias during testing. The tools were deployed in a high-risk context, which makes the failure more serious. This is not just a technical error; it is a civil rights issue.

Look at Northpointe's COMPAS tool from 2016. It showed large racial disparities in false-positive rates. It was used in sentencing decisions. Goldman Sachs had a similar issue with credit limits in 2019. Women received far lower limits than spouses. Both cases show that bias in training data leads to harmful outcomes. Another example is iTutorGroup in 2023. Recruiting software automatically rejected applicants over an age threshold. The EEOC settled with them.

To prevent this, we need fairness controls. We must test for disparate impact. The 4/5ths rule is a standard test to check if a tool has a disparate impact. We need explainability so we can see why a resume was rejected. We need human review for all high-stakes hiring decisions. This ensures that the AI does not make final calls on protected groups without a human checking the logic.

Under the EU AI Act, Article 14 requires human oversight and technical documentation. The act also bans high-risk AI systems that have been found to cause discrimination. We must document how we test for bias and what thresholds we use. We must also have a plan to override the AI if it makes a discriminatory decision.

More from our platforms

These sister platforms cover the parts of this problem that sit outside governance.

• Argus (argus.threatclaw.ai) records every trace an AI application produces and scans it for prompt injection, jailbreaks and data leaks, including the attacks hidden inside retrieved documents and tool results rather than in what the user typed. Governance decides what an AI agent is allowed to do. Argus shows what it actually did.
• ThreatClaw (www.threatclaw.ai) tracks the threat side of the same systems: 22 live intelligence feeds, exploitation predicted before it is officially confirmed, threat actor profiles, and detection rules you can deploy straight away. A control is only as good as the threat it is sized against.