Glossary
AI disclosure glossary
Plain definitions of the terms used across AI-content-disclosure law and the AIDisclose standard.
- ai-disclosure.json
- An open, machine-readable manifest served at a site's well-known path that declares which AI systems the site operates and how each is disclosed. It is the file the AIDisclose standard defines.
- Conformance level
- A measure of how mature a site's AI disclosure is, independent of any single law. AIDisclose defines three: Level 1 Declared, Level 2 Rendered, and Level 3 Evidenced.
- Level 1 Declared
- A valid ai-disclosure.json manifest is served at the site's well-known path. The site has stated which AI systems it operates.
- Level 2 Rendered
- Level 1 plus the declared disclosures are visible on the page: the interaction notice and the content labels appear where the manifest says they will.
- Level 3 Evidenced
- Level 2 plus a retained, hash-indexed evidence trail that records the disclosures were live on given dates, exportable as an audit-ready document.
- Provider
- The party that develops an AI system or places it on the market. Under the EU AI Act, providers carry the interaction-disclosure and synthetic-media marking duties.
- Deployer
- The party that uses an AI system under its own authority. Under the EU AI Act, deployers carry duties for emotion and biometric notices and for labeling deepfakes.
- Deepfake
- AI-generated or AI-manipulated image, audio, or video that resembles real people, objects, or events. Several laws require deepfakes to be disclosed.
- Machine-readable marking
- A disclosure embedded in a file or page so that software can detect it, for example provenance metadata or a watermark. The EU AI Act Article 50(2) requires it for synthetic media.
- Latent disclosure
- A machine-readable provenance signal embedded in AI-generated media. California SB 942 requires covered providers to embed latent disclosure in AI image, video, and audio.
- Visible disclosure
- A marker a person can see or hear, such as a label on an image or a spoken notice. Many laws pair it with a machine-readable marker.
- Synthetic media
- Image, audio, video, or text generated or substantially altered by AI. Synthetic-media labeling is the core of most AI-disclosure laws.
- C2PA Content Credentials
- An open provenance standard that attaches signed metadata about how a file was created or edited. AIDisclose aligns its machine-readable marking with C2PA.
- IPTC DigitalSourceType
- A controlled vocabulary for describing how media was produced, including AI generation. It is used to express AI provenance in image metadata.
- EU AI Act Article 50
- The transparency section of Regulation (EU) 2024/1689, covering AI-interaction disclosure, synthetic-media marking, emotion and biometric notices, and deepfake labeling. Enforceable from 2 August 2026.
- California SB 942
- The California AI Transparency Act, which requires large generative-AI providers to embed latent disclosure, offer a visible label, and provide a free detection tool for AI image, video, and audio. Enforceable from 2 August 2026.