On Thursday, the Wikimedia Foundation announced licensing agreements with Microsoft, Meta, Amazon, Perplexity, and Mistral AI, formalizing a system for major technology companies to compensate the nonprofit for using Wikipedia content to train their artificial intelligence models. These models power AI assistants like Microsoft Copilot and OpenAI's ChatGPT.

The deals mark a significant shift from the previous practice of these companies scraping Wikipedia for data without explicit permission. With these agreements, most major AI developers have now joined the Wikimedia Enterprise program, a commercial subsidiary that offers API access to Wikipedia's extensive database of 65 million articles. This access provides higher speeds and volumes of data compared to the free, public APIs. The Wikimedia Foundation did not disclose the specific financial terms of these new partnerships.

These new partners join Google, which established a similar agreement with Wikimedia Enterprise in 2022, along with smaller entities like Ecosia, Nomic, Pleias, ProRata, and Reef Media. The revenue generated from these licensing deals is intended to help offset the substantial infrastructure costs associated with maintaining Wikipedia. The nonprofit relies primarily on small public donations while its content has become a crucial resource for training AI models.

The use of Wikipedia content in AI training highlights the complex relationship between open-source knowledge and the rapidly evolving field of artificial intelligence. AI models, particularly large language models (LLMs), require massive datasets to learn and generate human-like text. Wikipedia, with its vast and collaboratively edited collection of articles, has become an invaluable source of information for these models.

The licensing agreements raise important questions about the ethical and economic implications of using publicly available data to train commercial AI systems. While Wikipedia benefits from the revenue generated by these deals, the broader implications for the future of open knowledge and AI development remain to be seen. The Wikimedia Foundation's move to monetize access to its data reflects a growing trend among content creators seeking compensation for the use of their work in the AI industry. The ongoing developments in this area could potentially reshape the landscape of AI training and data access in the years to come.