We introduce a new collection of spoken English audio suitable for training speech recognition systems under limited or no supervision. It is derived from open-source audio books from the LibriVox ...

A US court has ruled Anthropic was not breaching copyright rules when it trained its AI model on books.