What it means
Data classification is the practice of labelling information according to its sensitivity and protection needs, for example public, internal, confidential or restricted, so that appropriate rules apply. It determines what data can be used with which AI tools.
Real-world example
A federal agency publishes simple rules for staff: public information may be used with the approved cloud AI assistant; internal information only with the agency's internally hosted assistant; documents classified as "VS – Nur für den Dienstgebrauch" and personal data from case files may not be used with either until a dedicated solution is approved.
Why it matters for AI adoption
Clear classification makes AI rules understandable and usable for staff. Without it, people either avoid AI entirely out of uncertainty or use it with data they should not. Classification also guides architecture decisions, such as cloud versus on-premises hosting.
Questions to ask
- Which classes of data may be used with each of our AI tools?
- Do staff know these rules, and are they enforced technically where possible?