Databricks fails to shake authors' copyright claim
Briefly

Databricks fails to shake authors' copyright claim
"Judge Breyer stated, 'They directly tie their infringed works to DBRX, and the employee statements provide supporting inferences when read in context, particularly when viewed alongside other more direct statements.'"
"Databricks argued that the authors cannot prove DBRX was trained with the Book3 data, asserting that they have provided fourteen depositions and thousands of pages of documents."
Databricks is embroiled in a class action lawsuit from authors claiming its LLM, DBRX, was trained on pirated books, including 196,000 titles. Judge Charles Breyer denied Databricks' motion to dismiss, allowing the lawsuit to continue. The case revolves around the connection between DBRX and the MosaicLM model, which used the RedPajama dataset. Authors assert their works are directly linked to DBRX, supported by employee statements. Databricks has submitted extensive documentation to defend itself, but the judge seeks additional information.
Read at Theregister
Unable to calculate read time
[
|
]