Bottom line: Unknown and excessively distributed data endanger compliance, data security, and the quality of AI systems alike.
Up to 80 percent of enterprise data is considered dark data — information whose existence and location are unknown to organizations. This creates significant compliance and security risks while simultaneously undermining the quality of AI systems.
Dark data typically encompasses sensitive information such as personal data, strategy documents, contracts, customer lists, source code, and design drawings. Organizations can only fulfil their regulatory obligations if they know what data they own and where it is located. GDPR, for example, not only requires that data not be stored outside the EU, but also obligates organizations to provide information and delete data upon request — which is impossible without transparency.
A major operational problem is oversharing within organizations. Information is frequently shared with too many people in order not to forget anyone. Missing or restrictive official sharing tools lead to the formation of shadow IT, which organizations have no visibility into. Added to this: the use of AI tools such as Copilot or ChatGPT amplifies this phenomenon, as these systems need access to as many files as possible to be universally applicable. Typical sources of error also include forgotten or poorly configured cloud storage buckets — as in 2023 with Volkswagen subsidiary Cariad, where movement data from 800,000 electric vehicles were freely accessible on the internet.
From an AI perspective, dark data and oversharing are particularly problematic. AI systems that access excessively large and disorganized data volumes produce lower quality answers because they also process outdated or redundant information. Users can furthermore receive answers containing information not intended for them. Without regular data scans and continuous monitoring, dark data emerges anew, creating a data security risk and an obstacle to reliable AI deployment.
Source: www.it-daily.net · Published July 6, 2026
Lumi AI News — AI-assisted curation pursuant to Art. 50 EU AI Act. Paraphrase and classification through Lumi News Pipeline v1.7.3.