Discrimination Risk: where technical bias meets legally protected rights

Definition

Discrimination risk converts AI bias into legal liability. Zertia audits under ISO 42001 + EU AI Act Article 10 + fundamental rights.

Key facts

• Discrimination risk in AI is the legally protected harm that occurs when algorithmic bias produces disparate impact on protected groups.
• EU Charter of Fundamental Rights Article 21 and EU Equal Treatment Directives prohibit discrimination, including algorithmic forms.
• The Schufa CJEU ruling (2023) confirmed credit scoring algorithms fall within GDPR Article 22 and discrimination law scope.
• EU AI Act high-risk AI deployments in credit, recruitment, education and insurance face explicit anti-discrimination obligations.
• Zertia audits discrimination risk under ISO 42001 with explicit EU AI Act Article 10 and fundamental rights mapping.

Frequently asked questions

What is discrimination risk in AI?

The legally protected harm that occurs when algorithmic bias produces disparate impact on protected groups (race, gender, age, disability, sexual orientation). Discrimination risk converts bias from a technical defect into a binding legal liability under EU and Member State law.

How does discrimination risk differ from algorithmic bias?

Bias is a statistical property of the model output; discrimination risk is the legal consequence when that bias affects protected groups in regulated sectors (credit, employment, insurance, education, healthcare). Not all bias is discrimination, but all discrimination starts as bias.

How is discrimination risk audited?

Through ISO 42001 with explicit EU AI Act Article 10 mapping and fundamental rights assessment (ISO 42005). Zertia integrates protected-attribute analysis within audit scope.