Awesome-AIGC-Detection

本仓库由 OpenClaw (🦞) 和人类共同维护

---

🔍 A comprehensive collection of papers, datasets, and tools for AI-Generated Content (AIGC) detection research.

This repository tracks research on detecting AI-generated content across image, video, text, audio, and multimodal settings. Paper, dataset, tool, README, and timeline metadata are generated from structured files under data/.

Recent image/video entries are cross-referenced with ant-research/Awesome-AIGC-Image-Video-Detection.

🕐 Interactive Timeline · Local HTML · 📊 View as Table

---

🚀 Quick Start

git clone https://github.com/MuskAI/Awesome-AIGC-Detection.git
cd Awesome-AIGC-Detection

python3 scripts/validate_data.py
python3 scripts/generate_readme.py --check

---

📊 Statistics

By Venue

---

📋 Table of Contents

• News
• Datasets
• Research Navigator
• Papers
• Tools
• Adjacent Resources
• Contributing

📰 News

Fresh signals from platforms, policy, provenance, and industry deployment, with China-specific updates pulled into a dedicated watch lane.

Latest Signal

2026-05-18 · Industry / Identity

GetReal Security launches real-time identity verification and deepfake detection platform — PR Newswire / GetReal Security

Signal: GetReal announced general availability of GetReal Protect, combining multimodal deepfake detection with continuous identity verification for voice and video workflows.

Why it matters: AIGC detection is moving from offline forensic review into live enterprise authentication and incident response.

Reading The Trend

Coverage mix: Platform x3, China Policy x3, Industry x2, Policy x2, Provenance x1, China Industry x1, Research x1, China Platform x1. The current flow spans domestic governance and global deployment: platform-scale likeness search, capture-time provenance, labeling standards, review workflows, and real-time fraud defense.

China Watch

Domestic updates are especially useful for tracking how AIGC detection becomes platform policy, labeling infrastructure, rights protection, and video governance.

• 2026-04-30 · China Policy / AI Governance 中央网信办部署开展“清朗·整治AI应用乱象”专项行动 — 中国网信网 Signal: 中央网信办启动为期4个月的专项行动，重点整治生成合成内容标识落实不到位、AI数据投毒、AI换脸拟声滥用等问题。 Why it matters: 国内治理重点已经从单点内容处置扩展到源头备案、安全审核、标识互认和检测能力建设。

• 2026-04-08 · China Policy / Video Governance 国家广播电视总局通报“AI魔改”视频治理工作取得实效 — 国家广播电视总局 Signal: 广电总局通报专项治理结果，清理大量基于经典电视剧作品进行AI魔改的违规视频，并推动平台建立常态化治理机制。 Why it matters: AIGC视频检测在国内内容平台中正从研究问题变成审核、版权、文化安全和平台责任问题。

• 2026-04-02 · China Industry / Likeness Rights 中广联演员委员会就AI换脸合成、声纹克隆、影视素材魔改发布声明 — 澎湃新闻 Signal: 中国广电联合会演员委员会针对AI换脸、声纹克隆、影视素材魔改和未经授权采集训练数据等侵权行为发布声明。 Why it matters: 国内演艺行业把检测和溯源需求推向肖像权、声音权、训练数据授权和二创边界。

• 2026-02-12 · China Platform / Labeling 小红书要求AI生成合成内容主动标识，未标识内容将限制推荐 — 新浪科技 Signal: 小红书公告称将持续加强AI生成合成内容识别检测能力，未主动标识的AI生成合成内容将由平台添加标识并限制分发。 Why it matters: 国内社区平台正在把检测结果直接接入分发、补标、申诉和举报机制。

• 2025-09-01 · China Policy / Labeling Standard 《人工智能生成合成内容标识办法》正式施行 — 央视网 Signal: 国家网信办等四部门发布的标识办法正式施行，要求AI生成的文字、图片、音频、视频和虚拟场景等内容明确标识。 Why it matters: 国内检测生态有了制度入口，显式标识、隐式标识、元数据和传播端风险提示会影响后续工具与评测。

Global Field Notes

• 2026-05-15 · Platform / Likeness YouTube expands AI likeness detection to all adult users — The Verge Signal: The Verge reported that YouTube is expanding its likeness detection program to users aged 18 or older with a YouTube account. Why it matters: Consumer-scale enrollment turns deepfake discovery into a mainstream safety feature and raises new questions about false matches and review workflows.

• 2026-05-11 · Provenance / Camera Canon introduces C2PA-compliant Authenticity Imaging System for news organizations — Canon Global Signal: Canon announced a C2PA-based system for supported cameras that preserves provenance from capture through newsroom workflows. Why it matters: Secure capture and provenance metadata are becoming a practical complement to detector-only pipelines.

• 2026-04-30 · Industry / Fraud Sumsub launches Adaptive Deepfake Detector for fraud prevention — PR Newswire / Sumsub Signal: Sumsub released an upgraded deepfake detection system focused on fast model adaptation and multi-signal fraud analysis. Why it matters: The fraud setting emphasizes continual learning, liveness, device signals, and session context beyond media pixels alone.

• 2026-04-21 · Platform / Likeness YouTube expands likeness detection to the entertainment industry — YouTube Official Blog Signal: YouTube expanded its AI likeness detection access to talent agencies, management companies, and celebrities. Why it matters: Large platforms are operationalizing face-likeness detection as a rights and removal workflow, not only as a research benchmark.

• 2026-03-26 · Policy / Market UK DSIT publishes analysis of the deepfake detection technology market — GOV.UK Signal: The UK Department for Science, Innovation and Technology published a market and evidence review for deepfake detection technology. Why it matters: Policy interest is shifting toward supply, demand, barriers, and future growth of detection capabilities.

• 2026-03-10 · Platform / Public Interest YouTube expands likeness detection to civic leaders and journalists — YouTube Official Blog Signal: YouTube began piloting likeness detection access for government officials, journalists, and political candidates. Why it matters: High-risk public figures are a key deployment target for deepfake discovery, identity verification, and takedown workflows.

• 2026-02-19 · Research / Provenance Microsoft publishes study on media integrity and authentication methods — Microsoft Signal Signal: Microsoft discussed a media integrity report covering provenance, watermarking, fingerprinting, and their limitations. Why it matters: The field is converging on layered authenticity signals instead of relying on a single detector score.

• 2026-02-05 · Policy / Evaluation UK government announces deepfake detection evaluation framework — GOV.UK Signal: The UK government announced work with technology companies, academics, and experts to evaluate deepfake detection tools against real-world threats. Why it matters: Benchmarking is becoming a public-sector requirement, especially for fraud, abuse, impersonation, and safety use cases.

🧭 Research Navigator

This repository is curated as a map, not just a shelf: use the routes below to jump from problem type to representative papers, code, and benchmarks.

Fast Lanes

Research Tracks

Modality-Year Heatmap

2026 Watchlist

📚 Datasets

📝 Papers

🔧 Tools

Browser Extensions

Libraries & SDKs

Online Detection Tools

Provenance & Authentication

🧩 Adjacent Resources

Resources here are useful context for authenticity, misinformation, and provenance work, but they are intentionally kept out of the core AIGC detection paper list.

• 2026.05 · Multimodal fake news / misinformation detection The AI Slop Intelligence Dashboard Problem (James Sawyer Field Notes) Why adjacent: Related to authenticity and misinformation, but the primary task is critical analysis of AI-generated intelligence dashboard claims rather than direct AIGC content detection. Links: Resource

• 2025.04 · Multimodal fake news / misinformation detection Exploring Modality Disruption in Multimodal Fake News Detection (arxiv) Why adjacent: Related to authenticity and misinformation, but the primary task is not direct AIGC content detection. Links: Resource

• 2024.01 · Multimodal fake news / misinformation detection MiRAGeNews: Multimodal Realistic AI-Generated News Detection (ACL) Why adjacent: Related to authenticity and misinformation, but the primary task is not direct AIGC content detection. Links: Resource

🏙️ Others

• Paper backlog: PAPERS_TO_ADD.md
• Timeline source: docs/timeline_data.json, generated from data/papers.json
• Boundary resources: data/adjacent.json, generated into Adjacent Resources

---

🤝 Contributing

Contributions are welcome. To add or update entries:

1. Edit the structured files in data/.
2. Run python3 scripts/validate_data.py.
3. Run python3 scripts/generate_readme.py.
4. Open a Pull Request with the data and generated files together.

Please keep placeholder links such as https://arxiv.org/abs/ out of data/; leave incomplete candidates in PAPERS_TO_ADD.md until metadata is complete.

---

📬 Contact

If you have questions or suggestions, please open an issue.

---

Last generated on 2026.05.19