⚖️ OpenAI use of Copyrighted Content

Author

Liam Gill
April 17, 2025

In partnership with

Allegations Surface Around GPT-4o’s Use of Copyrighted Content

A new report from the AI Disclosures Project alleges that OpenAI’s GPT-4o model may have been trained on copyrighted, non-public content from O’Reilly Media without a licensing agreement. Using a method known as DE-COP — designed to detect whether a model has seen specific human-authored texts — the report found a high degree of “recognition” of O’Reilly’s paywalled book content by GPT-4o, far more than earlier models like GPT-3.5. While the report is cautious in its conclusions and acknowledges alternative explanations (such as user-provided excerpts), it adds fuel to ongoing legal and ethical concerns around how AI models are trained.

Prepare for your next business move with our International Hiring Guide

At Deel, we've simplified a world's worth of global hiring information. Our expert teams can help you navigate quick and compliant hiring in 150+ countries and so much more.

Legal Risk is Rising for Startups Using Unlicensed Training Data

For tech startups, particularly those building or training AI models, this development raises the stakes on data provenance. It’s no longer enough to rely on “fair use” arguments or vague interpretations of web scraping norms — regulatory scrutiny and potential litigation are becoming real risks, especially for models trained on unlicensed content. Founders should implement clear data sourcing policies, maintain logs of licensed and open datasets used in training, and ensure they understand the legal boundaries of synthetic data versus copyrighted material. Startups that proactively adopt transparent data governance may have a competitive edge, both in court and in the market.

A Market Opportunity in Licensing and AI Transparency

This controversy also presents an opportunity for startups offering “clean” datasets, licensing platforms, or AI auditing tools. As large labs come under pressure to demonstrate lawful and ethical model training practices, there's growing demand for third-party tools that verify data sourcing or assist companies in negotiating licenses. Moreover, startups building foundation models may need to consider how to make their models more auditable, especially as investors, partners, and regulators begin asking tougher questions about how these systems are built.

In addition to our newsletter we offer 60+ free legal templates for companies in the UK, Canada and the US. These include employment contracts, investment agreements and more

Newsletter supported by:

Get Over $6K of Notion Free with Unlimited AI

Running a startup is complex. That's why thousands of startups trust Notion as their connected workspace for managing projects, tracking fundraising, and team collaboration.

Apply now to get up to 6 months of Notion with unlimited AI free ($6,000+ value) to build and scale your company with one tool.