6 min read
Research Knowledge distillation attacks can transfer and forge the statistical watermarks embedded in LLM outputs, undermining EU AI Act content provenance requirements and attribution-based trust systems before they fully take effect.
c2pa