# AI training data copyright
> Source: https://legalcode.md/skills/legalcode-ai-training-data-copyright | [HTML version](https://legalcode.md/skills/legalcode-ai-training-data-copyright)

Analyze copyright implications of AI training data.

Analyze copyright implications of AI training data. Use when evaluating whether scraped or licensed content can be used to train AI/ML models, assessing legal exposure from bulk data acquisition, reviewing TDM (text and data mining) rights under EU DSM Directive Articles 3 and 4, analyzing US fair use defenses for model training, checking compliance with EU AI Act Article 53 training data transparency obligations, assessing opt-out mechanisms (robots.txt, TDM reservations), reviewing Creative Commons and data license compatibility with AI training, evaluating right of publicity risks in training datasets, analyzing DMCA Section 1202 copyright management information claims, or preparing a litigation risk assessment for an AI training dataset. Covers the NYT v. OpenAI and Getty Images v. Stability AI litigation landscape, EU AI Act GPAI provider obligations, US fair use four-factor analysis for training, web scraping legality under the CFAA and ToS law, and training data licensing strategies. Jurisdictions: United States and European Union, with coverage of cross-border scenarios.

This skill analyzes the copyright, contractual, and regulatory implications of using content to train AI and machine learning models. It produces a classified risk assessment with actionable recommendations for AI developers, model operators, data teams, and legal counsel. **Covers:** - US fair use four-factor analysis for AI training data (17 USC § 107) - EU DSM Directive TDM exceptions (Articles 3 and 4 of Directive 2019/790/EU) - EU AI Act GPAI provider transparency obligations (Article 53, Regulation (EU) 2024/1689) - Web scraping legality under the CFAA and Terms of Service law - robots.txt and machine-readable TDM reservation mechanisms - Creative Commons license analysis for training data compatibility - Right of publicity risks in training datasets (US state law) - DMCA Section 1202 copyright management information (CMI) claims - Key litigation landscape (NYT v. OpenAI; Getty Images v. Stability AI; Andersen v. Stability AI; Raw Story Media v. OpenAI; Kadrey v. Meta) [VERIFY current status] - Training data licensing strategies and best practices - Memorization/regurgitation risk in model outputs **Does not:** - Draft data licensing agreements (see `legalcode-technology-license-agreement`) - Assess database sui generis rights in isolation (addressed where relevant) - Handle patent or trademark issues for AI (see `legalcode-patent-landscape-analysis`) - Provide ...

## Jurisdictions

Global



---
*Legalcode -- legalcode.md*