Free Software Foundation calls for free-range LLMs rather than factory-farmed AI
F is for Free, FSF, and fat chance
UPDATED The Free Software Foundation (FSF) has rattled a saber at Anthropic over the use of its materials in training the AI vendor's models, urging it to set its LLMs free.
Along with many other copyright holders, the FSF received a settlement notice as part of the copyright infringement lawsuit Bartz v. Anthropic. In settling the class action lawsuit, Anthropic agreed in September to create a $1.5 billion fund to compensate authors whose works it had used to train its models without seeking or securing permission.
At the time, Anthropic managed to win on some aspects of the case early on – where tech-savvy judge William Alsup said it was "fair use" to use the books to train Large Language Models (LLMs), but left open the question of whether downloading them for that purpose was legal. Rather than wait for trial, the parties opted to settle. Copyright holders are now being contacted with offers of cash in lieu of potential damages.
One of the works found in the datasets used by Anthropic is Sam Williams's Free as in freedom: Richard Stallman's crusade for free software. According to the FSF, the book was published by O'Reilly and the FSF under the GNU Free Documentation License (GNU FDL).
"This," wrote the FSF, "is a free license allowing use of the work for any purpose without payment."
"Obviously, the right thing to do is protect computing freedom: share complete training inputs with every user of the LLM, together with the complete model, training configuration settings, and the accompanying software source code."
"Therefore, we urge Anthropic and other LLM developers that train models using huge datasets downloaded from the Internet to provide these LLMs to their users in freedom."
While users wait for hell to freeze over – the AI vendors are very unlikely to meet the FSF's demands (The Register asked Anthropic for comment) – the FSF noted it didn't have the resources for a protracted legal battle over the issue. It said, however, that if it were to participate in a lawsuit like Bartz v. Anthropic, and found its copyright and license violated, "We would certainly request user freedom as compensation."
It's laudable, but considering how much open source (code or otherwise) has already been slurped by AI vendors, this horse bolted long before anyone closed the stable door.
The finger-wagging is unlikely to move vendors, Anthropic included, unless a future action actually reaches trial before a settlement makes it disappear. ®
Updated to add at 1530 UTC:
A spokesperson for the Free Software Foundation (FSF) told The Register that its intention was "to highlight a broader principle: when large language models are developed using vast datasets drawn from the Internet, the best outcome for users is one that protects and expands user freedom."
It added: "The FSF publishes its works under licenses that guarantee the four essential freedoms: the freedom to use, study, modify, and share the work. Our concern is not about restricting use, but about ensuring that the development and distribution of these systems respect those same freedoms for all users."