AI
ETSI publishes a lifecycle standard for limiting deepfake misuse
ETSI TS 104 275, written by the institute's Securing Artificial Intelligence committee, sets safeguards across five stages of an AI system's life: secure design, development, deployment, maintenance and end of life. It asks developers for risk assessments, content filtering, labelling, continuous monitoring and human oversight, and supplies test scenarios with pass/fail criteria so protections can be demonstrated rather than promised. The work builds on ETSI's EN 304 223 baseline and supports EU AI Act goals on impersonation.