- Elon Musk's X tells users it could use their posts to train artificial intelligence models.
- Musk has launched his own AI startup and previously said it would use Twitter data to train it.
Elon Musk's X has changed its policies to allow information posted by users to be used to train its AI models.
X's new privacy policy, which is due to come into effect on September 29, states that the company "may use the information we collect and publicly available information to help train our machine learning or artificial intelligence models for the purposes outlined in this policy." This policy is not included in its previous terms, which are still posted online.
Musk responded to a post about this change on X, saying that it would only use publicly available information to train the AI and would not use "DMs or anything private."
During a live audio session on X – formerly Twitter – in July, Elon Musk said that his AI startup, xAI, would use public data from his social media platform to train its AI models. Insider reached out to X for comment but didn't immediately hear back. It is not clear how it will use the information from X and which AI models this relates to.
Musk has been critical of other platforms using X to train AI, threatening to sue Microsoft in April for "training illegally using Twitter data."
Elon Musk announced the launch of xAI, which he said would use artificial intelligence to "help understand the true nature of the universe," earlier this year. The launch came despite Musk previously warning that AI could lead to "civilization destruction," and signing a letter calling for a pause on the training of AI systems more powerful than OpenAI's GPT-4.
Musk's startup isn't his first foray into AI. He was an early backer of OpenAI but left the company's board in 2018. He previously told CNBC that he was crucial to the company's inception and even came up with its name.
He was also reportedly "furious" when ChatGPT made a splashy debut in November, unnamed sources told Semafor. And has since been critical of the chatbot, saying it is "trained to be politically correct, which is another way of saying untruthful things."
The technology that powers generative AI products such as ChatGPT, requires huge amounts of data to train. In the past, this data has been scraped from publicly available web sources such as Reddit and Wikipedia.
Some of these companies now want to cash in on this or stop it altogether. Reddit, for example, said in April it plans to charge AI companies for access to its data. While other sites such as The New York Times and Amazon are blocking OpenAI's web crawler.
Social media sites, with their endless supply of user-generated content, are a gold mine for AI companies hungry for new sources of training data.
Meta, which is also developing its own generative AI models, recently introduced an option for Facebook users to opt out of sharing their data to train its AI models. But this comes with limitations, per Gizmodo, and it's not guaranteed that all of a user's information will be removed from its AI training databases.