Council of Europe Launches Framework for Measuring Algorithmic Discrimination

The Council of Europe launched the European Assessment Methodology on Algorithmic Discrimination in September 2026, providing a structured framework for testing whether algorithmic systems produce discriminatory outcomes across protected characteristics including gender, race, age, and disability status. The methodology is designed to be usable by regulators, auditors, and organisations conducting internal reviews of their own AI systems, and draws on the Council of Europe's Convention 108+ on personal data protection as its legal foundation.

The framework establishes standardised testing protocols for detecting proxy discrimination, where an algorithmic system uses a facially neutral variable — such as postcode, internet browsing behaviour, or device type — to achieve an outcome correlated with a protected characteristic. Proxy discrimination is a recurring finding in regulatory investigations of algorithmic credit scoring, targeted advertising, recruitment AI, and content recommendation systems.

The methodology is not yet legally binding in any jurisdiction but is expected to influence the technical standards that national AI regulatory authorities and data protection authorities apply when assessing algorithmic systems for discriminatory effects. It aligns with the AI Act's requirements for bias testing in high-risk AI systems and with the GDPR's existing prohibition on processing that produces discriminatory outcomes through automated means.

Why it matters for business: A standardised discrimination testing methodology used by regulators changes the due diligence standard for companies deploying algorithmic systems in credit, employment, insurance, and consumer services. The methodology's focus on proxy discrimination — discrimination that operates through seemingly neutral variables — means that audit programmes designed only to check for direct use of protected characteristics will miss the patterns regulators are now trained to find. Internal AI audits conducted against the Council of Europe methodology provide a credible compliance record; those conducted against weaker proprietary frameworks do not.

Source: Council of Europe / Datalawgy, September 2026, https://datalawgy.substack.com/p/what-is-new-in-data-technology-and-008