AI training data controversy escalates: Another tech giant faces lawsuit over pirated books

[CryptoWorld] Recently, an interesting incident has occurred—another lawsuit in the tech industry over AI datasets. An author, Elizabeth Lyon, sued a well-known tech company, claiming that their large language model was trained using a dataset that included her copyrighted works.

What exactly happened? The issue centers around the SlimPajama-627B dataset. This dataset originates from the RedPajama project and includes a highly controversial collection called “Books3”—essentially a large amount of unlicensed book data. The company used this data to train the SlimLM AI model, and the author discovered that her works had been forcibly included.

This is not an isolated incident. Similar legal troubles are piling up, involving not only this company but also several other tech giants—accused of using protected content without authorization during AI development. This raises a core question: Can AI models freely use data from the internet and publications for training? How can the rights of content creators be protected?

From the perspective of Web3 and open-source communities, this incident reflects a larger contradiction. On one hand, AI development requires vast amounts of data; on the other hand, the rights of content creators must not be arbitrarily infringed. Finding a balance between the two has become a difficult challenge facing the entire tech industry. How these lawsuits will develop remains to be seen.