And when do Big Tech AI firms dress IP owners down for “overestimating” the value of their own creations?
Meta CEO Mark Zuckerberg has recently addressed the complex copyright issues surrounding data scraping for AI training — suggesting that most individual creators’ works are not valuable enough to impact the overall process.
In an interview with The Verge, Zuckerberg mentioned that Meta may form partnerships for valuable content but would rather walk away if payment is demanded, similar to its approach with news outlets.
His view is that creators often “overestimate the value of their content.” If creators object to their content being used, his firm would simply not use it, as that act would not significantly affect the outcome.
This stance comes amid the firm’s ongoing litigation over data scraping without permission. For instance, Meta was sued by authors, including Sarah Silverman, for allegedly using pirated copies of their work to train its Llama model. The case has not been favorable for the authors so far.
Zuckerberg has argued that unapproved data scraping should be allowed under US fair use law, asserting that copyright history involves determining control over published works, and fair use allows for transformation and building upon others’ creations without permission or compensation. However, some AI developers have interpreted fair use more broadly than most courts.
While the legal debate continues, several AI firms have started paid partnerships with major outlets. OpenAI has struck deals with news publishers and intellectual property (IP) owners such as Shutterstock, and Meta recently signed an agreement with Universal Music Group for AI-generated songs.
From the perspective of artists and creative, using “unofficial” tools to prevent their work from being used for AI training is possible, but terms of service on social media platforms often allow posted content to be used to train AI.
Zuckerberg has confirmed his firm trains its AI on public Instagram and Facebook posts, and that its future AI content strategy would likely mirror its own blunt response to proposed laws imposing fees for news links, typically — blocking entire genres of IP in such cases.