generative AI services
Systems creating novel content (text, images, audio, video) from learned patterns.
How the laws define this
Regulators agree that generative AI services encompass systems or technologies that produce synthetic or novel content—such as text, images, and audio—derived from prompts, user inputs, existing data, or learned patterns. They diverge, however, in their regulatory scope and emphasis. The Generative AI Service User Protection Guideline from the Broadcasting and Communications Commission focuses on operational boundaries, distinguishing pre-trained models, fine-tuning or continued learning, and runtime generation. Conversely, the MinCiencia instructive on AI use in government emphasizes data protection risks, differentiating generative platforms from other algorithmic tools and warning against uploading personal or sensitive data to uncontracted third-party services. Meanwhile, 人工智能生成内容标识管理办法 applies a broad scope that explicitly includes text, images, audio, video, and other digital assets generated through prompts or existing data.
Synthesised from the 3 statutory definitions below. Each one is quoted in full, with its source.
How jurisdictions differ
What is distinctive in each definition, binding law in force first.
| Jurisdiction | Instrument | Status | What is distinctive |
|---|---|---|---|
| China | Administrative Measures for the Labeling of AI-Generated Content | In Force | Broadly encompasses any services leveraging generative AI technology to produce text, images, audio, video, or other digital assets based on prompts or existing data. |
| South Korea | Generative AI Service User Protection Guideline | In Force | Defines in service-oriented language that clarifies boundaries between pre-trained models, fine-tuning/continued learning, and runtime generation. |
| Chile | MinCiencia instructive on AI use in government | In Force | Distinguishes generative models or services producing synthetic content from other algorithmic tools and cautions against uploading personal or sensitive data to uncontrolled third-party platforms. |
What it triggers
- In China, under Administrative Measures for the Labeling of AI-Generated Content, these services require explicit and implicit labeling for all AI-generated content.source
- In South Korea, the Generative AI Service User Protection Guideline sets four core principles and six implementation measures for protecting users of these services.source
- In Chile, the MinCiencia instructive on AI use in government provides operational standards for state agencies developing or procuring these services.source
Definitions (3)
Systems that create novel content such as text, images, audio, or video derived from learned patterns; defined in service‑oriented language that clarifies the operational boundary between pre‑trained models, fine‑tuning/continued learning, and runtime generation.
Models or services that produce synthetic content—such as text, images or audio—from user inputs; the circular distinguishes these from other algorithmic tools and cautions against uploading personal or sensitive data to uncontrolled third‑party generative platforms not contracted by the state.
Generative AI Services encompass any services that leverage generative artificial intelligence technology. These services are capable of producing various forms of content, including but not limited to text, images, audio, video, or other digital assets, often based on prompts or existing data.
Not to be confused with
Refers specifically to the AI technology that produces synthetic content, whereas this term covers services using that technology.
Specifies internet-accessible systems producing content from user prompts, while this term more broadly covers services or models.
Focuses on AI systems like foundation models capable of producing synthetic content, differing from this term's service or model orientation.
Questions and answers
What labeling is required for generative AI services in China?
China's Administrative Measures for the Labeling of AI-Generated Content mandate explicit and implicit labeling for all AI-generated content to ensure transparency and prevent misinformation. These rules apply to services producing text, images, audio, video, or other digital assets.
What user protections does South Korea recommend for generative AI services?
The Generative AI Service User Protection Guideline sets four core principles and six implementation measures to protect users of generative AI services producing text, audio, or images. It targets developers and clarifies operational boundaries between pre-trained models, fine-tuning, and runtime generation.
How does Chile's government guideline treat generative AI services?
The MinCiencia instructive provides non-binding operational standards for state agencies developing or procuring generative AI services that produce synthetic content from user inputs. It distinguishes these from other algorithmic tools and cautions against uploading personal or sensitive data to uncontrolled third-party platforms.
Related Terms
Generative Artificial Intelligence (Generative AI)
AI systems that produce new content (text, images, audio, video)....
Generative artificial intelligence
AI that produces synthetic content emulating its training data....
Generative AI (Gen AI)
AI capable of creating new content like text, images, video, or audio by learning from data....
Generative AI tools
Internet-accessible systems producing content from user prompts....
Generative Artificial Intelligence System
An AI system, often a foundation model, capable of producing synthetic digital content like audio, visual, or text....