By Nan Gao — AI & Tech Lawyer · CS Background 作者：高楠 — AI合规与科技法律师 · 计算机背景
A practical guide to navigating China's AI regulatory landscape — covering algorithm filing, generative AI service registration, content labeling compliance, and AI data governance.
面向 AI 开发者、创业者和产品团队的中国 AI 合规实操指南，涵盖算法备案、生成式 AI 服务备案与登记、AI 生成内容标识合规以及数据合规等核心问题。
⭐ If this project helps you, please consider starring the repo. ⭐ 如果这个项目对你有帮助，欢迎点一个 Star 支持。
- AI startup founders and co-founders / AI 创业公司创始人
- Developers building AI-powered products / 开发 AI 产品的工程师
- Product managers shipping AI features / 负责 AI 功能的产品经理
- In-house legal and compliance teams at AI companies / AI 公司的法务与合规团队
- International companies entering the China market with AI products / 携 AI 产品进入中国市场的海外公司
If you are building an AI product in China, this repository helps you quickly answer:
如果你正在开发或运营 AI 产品，这个仓库可以帮助你快速回答以下问题：
- Do I need to file my algorithm with regulators? / 我的产品是否需要进行算法备案？
- Does my product require model filing or registration? / 我的 AI 产品需要大模型备案还是登记？
- I'm building on a third-party model's API — what do I need to do? / 我调用别人的大模型API做产品，需要做什么？
- What labeling requirements apply to AI-generated content? / AI 生成内容需要满足哪些标识要求？
- What data compliance obligations apply to AI systems? / AI 产品涉及哪些数据合规义务？
- What should I know before taking my AI product overseas? / AI 产品出海前需要了解哪些合规问题？
When does compliance kick in? / 什么时候需要合规？
As soon as your AI product or feature is accessible to the public in China — whether through an app store, a website, a mini-program, a public API, or any other channel. There is no "too small to regulate" threshold. If it's public-facing and uses AI, the rules apply.
只要你的 AI 产品或功能在中国境内向公众开放——无论是通过应用商店、网站、小程序、公开API还是其他渠道——合规义务就已经触发了。没有"用户量太小不用管"的豁免。面向公众 + 使用了AI = 需要合规。
Pick your scenario / 找到你的场景：
1️⃣ Using algorithms but NOT generative AI / 使用算法但不涉及生成式AI
Your product uses recommendation, ranking, search/filtering, or scheduling algorithms to serve users, but does not generate content. 你的产品使用了推荐、排序、检索过滤或调度决策等算法技术为用户提供服务，但不涉及内容生成。
→ Algorithm Filing Guide / 算法备案指南
2️⃣ Built your own generative AI model / 自研生成式AI模型
You trained your own foundation model and provide generative AI services (text, image, audio, video) to the public. 你自己训练了基础模型，并面向公众提供生成式AI服务（文本、图片、音频、视频）。
→ Algorithm Filing算法备案 + Model Filing大模型备案
3️⃣ Building on a third-party model's API / 调用第三方大模型API开发产品
You call an existing model's API (e.g. DeepSeek, Qwen, Doubao, GPT via Azure) and build your own product on top. This is the most common scenario for AI startups. 你调用已有大模型的API（如DeepSeek、通义千问、豆包、通过Azure调用GPT等）来开发自己的产品。这是大多数AI创业公司的情况。
→ Algorithm Filing算法备案 + Model Registration大模型登记
4️⃣ AI-generated content / AI 生成内容
Your product generates or synthesizes text, images, audio, video, or virtual scenes. Since Sept 2025, all AI-generated content must carry explicit and implicit labels. 你的产品生成或合成文本、图片、音频、视频或虚拟场景。自2025年9月起，所有AI生成内容必须添加显式标识和隐式标识。
→ Content Labeling Compliance / AI 内容标识合规
5️⃣ AI training data / AI 训练数据
You collect, purchase, or process data to train or fine-tune models. Obligations include data source legality, personal information protection, and cross-border transfer rules. 你为训练或微调模型而收集、采购或处理数据。涉及数据来源合法性、个人信息保护、数据出境等义务。
→ AI Data Compliance / AI 数据合规
6️⃣ Going global / AI 产品出海
You're taking your AI product to overseas markets, or you're an international company entering China. Key issues include technology export controls, data transfer, and US/EU regulatory differences. 你正在将AI产品推向海外市场，或者你是进入中国市场的海外公司。核心问题包括技术出口管制、数据出境、中美欧法规差异。
→ Cross-border Compliance / 出海合规
⚠️ These are not mutually exclusive — for example, a GenAI product likely needs 2️⃣ or 3️⃣ AND 4️⃣.⚠️ 以上不互斥——比如一个生成式AI产品大概率同时需要 2️⃣ 或 3️⃣ 加上 4️⃣。
Still not sure? / 还是不确定？ See the Decision Flowchart / 判断流程图.
China's AI regulations are evolving rapidly. However, most official regulatory materials are:
中国的 AI 监管体系发展非常快，但目前大多数官方材料：
- scattered across multiple government websites / 分散在多个政府网站
- written in dense legal language / 使用较为复杂的法律语言
- rarely explained for developers and founders / 很少针对开发者和创业者进行解释
- almost never available in English / 几乎没有英文版本
This repository aims to organize these regulations into clear, practical compliance guidance for people actually building AI systems — in both Chinese and English.
本仓库尝试将这些法规整理成清晰、可操作的合规指南，用中英双语帮助真正开发 AI 产品的人理解和遵守相关要求。
China's AI governance is built on a layered system — foundational laws at the top, sector-specific regulations in the middle, and technical standards at the bottom:
中国的 AI 治理建立在分层体系上——顶层是基础法律，中间是部门规章，底层是技术标准：
基本法律
└─ 网络安全法 / 数据安全法 / 个人信息保护法
└─ 部门规章
├─ 互联网信息服务算法推荐管理规定 (2022.03)
├─ 互联网信息服务深度合成管理规定 (2023.01)
├─ 生成式人工智能服务管理暂行办法 (2023.08)
└─ 人工智能生成合成内容标识办法 (2025.09)
└─ 强制性国家标准 & 技术文件
├─ GB 45438-2025 AI生成合成内容标识方法
└─ 实践指南（服务提供者编码规则等）
📄 Detailed breakdown / 详细解读: /docs/regulatory-overview.md
Ready-to-use reference materials (anonymized):
可直接参考使用的模板（已脱敏）：
| 模板 Template | 说明 Description | 状态 Status |
|---|---|---|
| 算法备案材料清单 | Complete materials list with filling guidance / 完整材料清单及填写说明 | 🚧 |
| 安全自评估报告大纲 | Outline for GenAI security self-assessment report / 安全自评估报告大纲 | 🚧 |
| AI标识自查清单 | Self-audit checklist for content labeling / 内容标识合规自查清单 | 🚧 |
| 隐私政策AI条款模板 | AI-specific clauses for privacy policies / 隐私政策AI专项条款 | 🚧 |
| 用户协议AI条款模板 | AI-specific clauses for terms of service / 用户协议AI专项条款 | 🚧 |
| 数据出境评估要点 | Key points for data export assessment / 数据出境评估要点 | 🚧 |
Latest updates on China's AI regulatory developments, updated regularly.
中国 AI 监管政策动态，持续更新。
| 日期 Date | 事件 Event | 影响 Impact | 链接 Link |
|---|---|---|---|
| 2025.09.01 | AI内容标识办法正式施行 AI Content Labeling Rules effective | 所有AI服务须添加内容标识 All AI services must label content | 详情 |
| 2025.03.14 | 标识办法 + GB 45438-2025 发布 Rules + standard published | 6个月过渡期开始 6-month transition begins | 详情 |
| ... | ... | ... | ... |
📄 Full tracker / 完整追踪: /tracker/README.md
Nan Gao / 高楠
I'm a practicing lawyer in China focused on AI compliance and tech law, with a bachelor's degree in computer science and years of experience in the tech industry. I built this project because I believe AI compliance shouldn't be a black box — developers deserve clear, practical guidance, not just legal jargon behind a paywall.
我是一名中国执业律师，专注于 AI 合规与科技法方向。本科计算机专业，曾在互联网科技公司工作多年。做这个项目是因为我认为 AI 合规不应该是一个黑箱——开发者需要的是清晰、实用的指南，而不是藏在付费墙后面的法律术语。
California Bar Candidate (July 2026) · Bridging China-US AI regulatory practice
- 💼 LinkedIn:
- 📱 小红书 / 抖音:
- 📧 Contact:
Need professional help? / 需要专业服务？ This repo is for education and reference. If your company needs tailored compliance support — algorithm filing, GenAI registration, content labeling, or cross-border matters — feel free to reach out. 本仓库内容仅供学习参考。如需算法备案、大模型备案、标识合规或出海合规的专业法律服务，欢迎联系。
This repository is for informational and educational purposes only. It does not constitute legal advice. Regulations change frequently — always verify with official sources and consult qualified legal professionals for your specific situation.
本仓库内容仅供参考和学习，不构成法律意见。法规更新频繁，请以官方发布为准，具体问题请咨询专业律师。
© Nan Gao. All rights reserved.
You are welcome to read, reference, and share links to this repository. However, republishing, reproducing, or redistributing the content in whole or in part — including in articles, courses, paid products, or other repositories — is not permitted without prior written permission.
本仓库内容版权所有。欢迎阅读、引用和分享链接，但未经书面许可，不得全文转载、复制或再分发本仓库内容（包括但不限于用于文章、课程、付费产品或其他仓库）。
For licensing inquiries / 授权咨询, please contact
