Protected works are creative materials covered by copyright or similar rights, such as books, images, music, software, and design assets. In cyber security and AI governance, the term matters because these assets may be stored, copied, indexed, or generated inside digital systems without clear permission.
Risk appears when a model is trained on protected works, when stolen or scraped content is used in a pipeline, or when outputs closely resemble copyrighted material. Defenders reduce exposure with data provenance tracking, license checks, content review, and policies that limit reuse of third-party material. Attackers may also exploit weak controls by bulk scraping or injecting copyrighted content into training sets, creating legal and operational problems. In practice, protected works are a compliance boundary: if a system can ingest, transform, or reproduce them, security teams need controls for authorization, auditability, and output screening.