What happened
On 3 September 2026, ETSI published Technical Report TR 104 180 'Data Solutions; Development and identification of Data Quality Metrics', defining 18 standardised, mathematically-specified metrics (spanning fundamental quality, usability, fairness, and privacy/responsible-data-use dimensions) for quantitatively assessing whether datasets are fit for purpose in digital and AI ecosystems. The TR was validated via an open-source Data Quality Validation System proof-of-concept applied to industrial IoT sensor data and demographic data, developed with Sejong University, Daejeon University, South Korea's TTA, Italy's CNIT, and France's EGM. Confirmed via ETSI's own press release with json-ld datePublished=2026-09-03.
Why it matters
Data quality directly underpins AI security and trustworthiness — poor-quality, biased, or unrepresentative training/inference data is a root cause of model integrity failures and a vector adjacent to data-poisoning risk. TR 104 180 gives organisations (and regulators referencing 'data governance' obligations, e.g. under the EU AI Act) a standardised, reproducible way to self-verify and report dataset fitness, complementing existing AI security frameworks (OWASP AI Exchange, ISO/IEC 27090, prEN 18282) that assume — but don't measure — data quality as an input.
Action needed
AI/ML data engineering and governance teams should evaluate the 18 metrics for inclusion in data-intake and model-training pipeline quality gates; standards/compliance teams tracking EU AI Act data-governance provisions should map this TR into their AI Act Annex documentation toolkit as supporting evidence.