Reported information
ETSI sets out deepfake prevention requirements from system design
ETSI publicized specification TS 104 275 on 29 October 2026 to reduce harmful deepfakes produced by generative AI. The September-dated document supplements its cybersecurity requirements for AI models and systems.
It calls for pre-deployment risk assessment, human oversight, suitable filters and labeling, followed by ongoing monitoring. Test scenarios and pass/fail criteria are intended to document whether safeguards are implemented.
The stated aim is to intervene from the design stage, beyond detecting fake content after distribution. This is a technical specification: its presentation neither creates a new European law nor proves that products on the market meet its requirements.