Google Docs users can breathe a sigh of relief. Their documents are not as public as previously feared, with the tech giant reportedly confirming that they are not considered "publicly available" for AI training.
What Happened: Earlier this month, a report revealed that Alphabet Inc.'s
The representative for Google said that changing the share settings to "anyone with the link" does not make a document "public" for AI training purposes. To be considered "publicly available," a document must be posted on a website or shared on social media, where it can be found by a web crawler.
Documents shared privately, such as through email, are not used for AI training, said the Google representative. So, unless a user has publicly posted links to their Google Docs, the documents were probably not used for AI training or considered "publicly available."
Why It Matters: This statement from Sundar Pichai-led tech giant comes amidst a series of controversies around AI training data sources. Last month, Tesla and SpaceX CEO Elon Musk accused ChatGPT-parent OpenAI of stealing "everything" after questions about its data sourcing for training its text-to-video AI model, Sora, emerged.
OpenAI reportedly used more than a million hours of YouTube videos to train its latest AI model. YouTube CEO Neal Mohan previously suggested that if OpenAI used YouTube videos to train its AI model, Sora, it would be a clear violation of the platform's policies.
When the interviewer asked Mohan if Google was using YouTube data to train Gemini, the CEO said, "Google uses YouTube content really in accordance. Again back with those terms of service or individual contracts that we might have with creators or uploaders to our platform."
Meanwhile, in February earlier this year, Google also inked a $60 million annual deal with Reddit to enhance its AI capabilities.
