Technical

Deep synthesis technology

Generative/synthetic algorithms that create or edit network information.

How the laws define this

Both regulatory instruments agree that deep synthesis technology involves artificial intelligence methods, specifically deep learning, to generate, create, manipulate, or edit content. They share a core framework establishing that this technology covers both generating new material and altering existing data. The primary point of divergence lies in their scope and specific technical applications. Under the [Provisions on the Administration of Deep Synthesis of Internet-based Information Services], the focus spans widely across network information, explicitly detailing applications like text generation, text-to-speech, music generation, image enhancement, 3D reconstruction, virtual reality, and immersive virtual scenarios alongside face generation or replacement. Conversely, [人工智能生成内容标识管理办法] centers its scope around artificial intelligence or machine learning used to generate or manipulate content, placing explicit emphasis on biometric information such as expressions, voices, or faces, and explicitly highlighting technologies like deepfakes.

Synthesised from the 2 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	Provisions on the Administration of Deep Synthesis of Internet-based Information Services	In Force	This definition expands beyond algorithms to include related technologies like virtual reality and generative/synthetic methods, and lists specific examples such as text-to-speech, face replacement, image enhancement, 3D reconstruction, and immersive virtual scenarios.
China	Administrative Measures for the Labeling of AI-Generated Content	In Force	This definition narrows to artificial intelligence-based technology like deep learning or machine learning, with a focus on generating or manipulating biometric information such as expressions, voices, or faces, and explicitly associates it with deepfakes.

What it triggers

In China, under Provisions on the Administration of Deep Synthesis of Internet-based Information Services, it mandates real-name registration, content labeling, and algorithm filings for deep synthesis providers and technical supporters.source
In China, under Administrative Measures for the Labeling of AI-Generated Content, it mandates explicit and implicit labeling for all AI-generated content.source

Definitions (2)

Algorithms and related technologies (including deep learning, virtual reality and other generative/synthetic methods) used to create or materially edit network information such as text generation, text-to-speech, music generation, face generation/replacement, image enhancement, 3D reconstruction and immersive virtual scenarios.

Provisions on the Administration of Deep Synthesis of Internet-based Information ServicesDefinition 1 of 2

Deep Synthesis Technology refers to artificial intelligence-based technology, such as deep learning or machine learning, that is utilized to generate or manipulate content. This includes the creation or alteration of biometric information like expressions, voices, or faces, commonly associated with technologies such as deepfakes.

人工智能生成内容标识管理办法Definition 2 of 2

Not to be confused with

Generative artificial intelligence

Generative artificial intelligence produces synthetic content emulating its training data, while deep synthesis technology specifically creates or edits network information using generative/synthetic algorithms or AI methods.

Generative Artificial Intelligence System

A generative artificial intelligence system is an AI system like a foundation model capable of producing synthetic digital content, whereas deep synthesis technology refers to the algorithms or AI technologies used to generate or manipulate such content, often biometric.

Questions and answers

What obligations apply to deep synthesis providers in China?

Providers and technical supporters must undertake real-name registration, content labeling, and algorithm filings. The rules also prohibit harmful synthetic content.

What labeling is required for AI-generated content in China?

Explicit and implicit labeling is mandated for all AI-generated content to ensure transparency and prevent misinformation.

Does deep synthesis include biometric manipulation like deepfakes?

Yes, it includes the creation or alteration of biometric information like expressions, voices, or faces, commonly associated with deepfakes.