Photobucket in negotiations to licence 13 billion images to AI companies
Image and video hosting website Photobucket is reportedly in talks with a number of generative artificial intelligence companies to license its database of 13 billion images and videos for the purpose of training text-to-image algorithms.
Photobucket once trumpeted more than 70 million users, but that number has dwindled to just 2 million in recent years. As a result, the company, which employs 40 people, has increasingly looked for alternative revenue streams.
In an interview with Reuters, CEO Ted Leonard said Photobucket is in talks with "multiple tech companies" to license its database of material, with rates of between 5 cents and $1 US dollar per photo and more than $1 per video being discussed.
Leonard says he can offer the material because last October the company changed its terms of service to grant it “unrestricted rights” to sell any and all uploaded content for the purposes of AI training.
“We need to pay our bills, and this could give us the ability to continue to support free accounts,” Leonard told Reuters.
Tech giants like Google, Meta and Microsoft-backed have used reams of data scraped from the internet for free to train generative AI models like ChatGPT. The companies say the technology would be cost-prohibitive if they couldn't use vast archives of free scraped web page data, which they describe as "publicly available."
The companies also maintain this is both legal and ethical, but as the tools require seemingly endless information to train their algorithms, image and video databases, even those privately held, have become a valuable proposition.
According to Reuters, in the months after ChatGPT debuted in late 2022, for instance, companies including Meta, Google, Amazon and Apple all struck agreements with stock image database Shutterstock to use hundreds of millions of images, videos and music files in its library for training. These deals ranged from $25 million to $50 million each.
Cover image: Wikimedia Commons