<![CDATA[Generative artificial intelligence enables ordinary users and organised actors to create text, images, audio, and video that can imitate real persons, institutions, and events at low cost. In Indonesia, this capability intersects with existing problems of hoaxes, impersonation, consumer deception, and personal-data misuse. This article argues that Indonesia needs a binding obligation to label AI-generated and AI-manipulated content, especially when such content is distributed to the public, used in electronic transactions, or capable of affecting public trust. The research uses a normative juridical method with statutory, conceptual, and comparative approaches. It examines the Electronic Information and Transactions Law, the Personal Data Protection Law, rules on electronic system providers, and Indonesia's AI ethics policy, then compares them with transparency models in the EU AI Act, C2PA provenance standards, NIST guidance, and Chinese synthetic-content labelling rules. The analysis finds that Indonesian law can sanction false information, fraud, and unlawful data processing after harm occurs, but it does not yet impose a clear ex ante duty to disclose synthetic origin. A labelling regime would close this gap without turning every synthetic work into prohibited content. The article proposes a layered model: creator-side disclosure, provider-side machine-readable provenance, platform-side label preservation, heightened duties for political and financial-risk content, and due-process safeguards for lawful expression. Mandatory labelling should therefore be framed as a transparency and consumer-protection obligation, not as a blanket restriction on innovation]]>
