Anthropic has doubled down on its opposition to major music publishers’ sought injunction. Photo Credit: Igor Omilaev
Anthropic is doubling down on its opposition to a preliminary injunction push from major music publishers, which say the AI giant’s Claude chatbot has infringed on their protected works.
Amazon-backed Anthropic just recently refiled its opposition to the plaintiff publishers’ desired preliminary injunction, on the heels of a court order granting and denying in part jointly sought redactions.
Now, those redactions (impacting a relatively small portion of Anthropic’s argument, albeit including interesting details like how much it would cost to comply with the injunctions) are reflected in the appropriate documents.
For a bit of quick background, it’s now been more than a year since Concord, Universal Music, and others first requested said relief. One of the desired injunctions pertains to removing the plaintiffs’ protected works from the generative AI system’s training data, with the other calling for an order blocking protected lyrics from appearing in Claude outputs.
At the top level, Anthropic in opposing the injunctions largely reiterated longstanding arguments – chief among them that using copyrighted works to train LLMs constitutes fair use.
“While Anthropic is confident that using copyrighted content as training data for an LLM is a fair use under the law—meaning that it is not infringement at all—there is no basis to conclude that money damages would not make Plaintiffs whole if they ultimately prevailed on the merits,” the AI platform indicated.
“The Court, accordingly, should not stretch on this underdeveloped record to get ahead of the other cases where the analogous substantive copyright issue will be adjudicated on a full summary judgment or trial record,” Anthropic continued.
Unsurprisingly, Anthropic also reiterated at length the volume of data behind Clyde – the chatbot “learns the patterns of language from trillions of tiny textual data points” – and explored the training specifics, probably including “some copyrighted works,” at hand.
Adjacent to the point, the AI developer stressed the now-common transformative argument. Utilizing “copyrighted song lyrics as part of a multi-trillion token dataset to train a generative AI model about the world and how language works is the very definition of ‘transformative’ under the fair use doctrine,” per the defendant.
Furthermore, certain cited research details “preceded the commercial release of Claude by nearly a year,” according to the legal text. And since then, Anthropic has purportedly added “a broad array of safeguards to prevent reproduction of copyrighted works from occurring in Claude outputs.”
All told, those safeguards mean there’s “no reasonable expectation that the complained-of activity is likely to recur,” according to Anthropic, which also refuted allegations of ongoing market and licensing harm stemming from the purported infringement.
Lastly, in a separate declaration in support of the injunction opposition, Anthropic co-founder Jared Kaplan rattled off his credentials and explained at length Claude’s training particulars.
Heading into the new year, it’ll be worth closely monitoring the status of the injunction battle – not to mention the overarching infringement dispute, a substantial portion of which might be tossed in the not-so-distant future.