Legal AI for all

January 27, 2026

AI works by accumulating data coming from existing sources. The sources can be anything on the Internet, social media posting, etc. Anthropic has observed that more reliable information comes from published books, and therefore are using published books as sources to provide more reliable information to their engine named Claude.

The idea proves to be fruitful. However companies bypass the fact that many book contents is copyrighted. Those companies do not necessarily have the right to exploit this information legally and can be sued.

The article published in the Washington Post shows how some companies are using books to acquire their source data:

"How Silicon Valley built AI: Buying, scanning and discarding millions of books", by Aaron Schaffer, Will Oremus and Nitasha Tiku, Washington Post , January 27, 2026

https://www.washingtonpost.com/technology/2026/01/27/anthropic-ai-scan-destroy-books

Therefore, it would be useful to build an AI technology with full disclosure of the legal agreements that allow sources to be used. Respecting the law would create hurdles, slow-downs, and limitations but would ensure that information acquired by the AI product is duly authorized. Furthermore, if the sources are published, it may help the users figuring out whether they can consider that this information can be trusted.

Along these lines, we can envision building an AI toolkit that would have all the computing processes ready to run, and could be populated by custom sources. That would allow any company, organization, research center, etc., to build their own AI agent that would process only information that they own, and therefore would be prone to errors or hallucinations.

We can also start thinking how to amend or adapt the copyright laws for explicitly allowing -- or restricting -- the use of AI agents. At minimum, the exploitation as sources for AI engines should be listed in the contracts, alongside with the rights to exploit the contents for audiovisual media, translations in other languages, etc.