Data lineage remains urgent despite EU AI Act delay

The delayed compliance deadline for high-risk AI systems under the EU AI Act does not reduce the need for stronger data governance. The law requires organizations using AI in areas such as credit scoring, insurance underwriting, and hiring to prove that training data is traceable, well governed, and assessed for bias.

Most enterprise data infrastructure was not built for that level of visibility. GDPR focused on data storage and access, while the AI Act requires organizations to trace data from its original source through transformations and into model outputs. AI model validation can take between nine and 12 months when lineage infrastructure is already in place.

Financial services face particular exposure because models trained on historical data can reproduce past biases in automated decisions. Existing obligations under BCBS 239 already require banks to demonstrate data accuracy, integrity, and risk data aggregation, making the AI Act an intensification of established governance demands rather than a wholly new challenge.

Data lineage is positioned as core infrastructure rather than compliance overhead. Bi-temporal lineage can recreate the data state used for model training and help teams simulate the downstream effects of data or schema changes before they reach production.