Meta has confirmed plans to train AI models using European users’ public content and conversations with its Meta AI chatbot.
Last year, Meta announced plans to train its AI on user data, but pressure from activists and regulators, as well as tighter data privacy laws in the EU, led it to pause that plan indefinitely. Meta also held back the rollout of its Meta AI assistant in Europe until last month.
A year on, Meta said in a blog post that it would soon restart that training, using interactions that people have with its AI systems such as questions and queries, as well as “public content shared by adults” on platforms including Facebook and Instagram.
However, Europeans will be able to object to their public data being used to train AI.
The company said the aim was to help better reflect European culture, history, and languages, suggesting there was some value to users handing over their data for training.
“We believe we have a responsibility to build AI that’s not just available to Europeans, but is actually built for them,” Meta said in its blog post.
“That’s why it’s so important for our generative AI models to be trained on a variety of data so they can understand the incredible and diverse nuances and complexities that make up European communities.”
In its blog post, the tech giant added this spans “everything from dialects and colloquialisms, to hyper-local knowledge and the distinct ways different countries use humor and sarcasm on our products”.
Drawing upon data sourced from these domains will be vital for the company’s efforts to develop more advanced, multi-modal AI models, which harness text, voice, video content, and imagery.
Leading AI developers have faced repeated criticism for using ebooks, online articles, and other digital content scraped from the web to feed the growing data needs of their models, with some facing lawsuits.
Indeed, Meta noted that this type of AI training wasn’t unique to the firm, and is how it and other AI developers train their generative models.
Meta insisted it has “engaged constructively” with regulators on the use of public posts for AI training.
“Last year, we delayed training our large language models using public content while regulators clarified legal requirements,” Meta said. “We welcome the opinion provided by the EDPB in December, which affirmed that our original approach met our legal obligations.”
Meta initially paused the use of such data in the UK, but went ahead with the plans after it changed its approach to satisfy demands from the ICO.
Europeans using Meta’s platforms will receive a notification in apps and via email that will lay out what data will be used and how it will help Meta train AI.
That message will also include a link to a form where people can object to their data being used to train AI, with Meta saying users can file the objection now before the data scraping begins or “at any time” later if they change their mind.
When the plans were first announced last year, Meta offered an opt out form which should still be valid.
“We have made this objection form easy to find, read, and use, and we’ll honor all objection forms we have already received, as well as newly submitted ones,” Meta said.
Meta added that anyone listed as being below the age of 18 won’t have their data used for AI training, and stressed only public posts will be used.
“As we’ve previously mentioned, we do not use people’s private messages with friends and family to train our generative AI models,” Meta said.
