Measured on 105 synthetic test cases
Pattern matching removed 57% of personal details. Adding a trained model removed 98.7%.
Keeping personal data out of AI prompts
What ends up in prompts, how well each detection method removes it, what placeholders achieve under the September 2025 judgment of the Court of Justice, and what you still have to do.
Share of annotated personal details removed from 105 held-out synthetic test cases, measured on 15 September 2026. This does not measure whole-document anonymity. The guide explains the method and its limits.