How to ask Facebook's Meta to not train its AI models on some of your personal info
Hey, stop Zucking up my data!
Netizens can ask Meta, the home of Facebook, Instagram, and Threads, to not train its generative AI models on at least some of their personal data.
Like pretty much all businesses developing AI systems capable of producing text, images, and more, the social media giant scrapes the internet for material to teach various models. Folks can now ask Meta, via this online form, to remove some of their personal information from that training data to prevent it from being ingested by these models, if they qualify.
"Depending on where people live, they may be able to exercise their data subject rights and object to certain data being used to train our AI models. They can submit an objection form to us through the Privacy Center link," a Meta spokesperson told The Register on Thursday.
That form can be used to ask Meta to delete, inspect, or edit third-party-sourced information about yourself that may be fed into a neural network to train it.
For example, a person's name, details about their work, or contact information described in a public blog post could be swept up in data scraped by Meta for training, data that a model could late regurgitate to other people. If you don't want this info to be used to train the mega-corp's generative AI models, you can thus ask for it to be deleted.
Unfortunately, the new policy only covers third-party sources and doesn't extend to any personal data uploaded to Meta's social media platforms. In other words, Meta can use the text contained in posts or comments, or selfies and photos submitted by users to Facebook or Instagram, to train its AI models. People just have a little more control over the data that comes from outside Meta's empire.
"Since it takes such a large amount of data to teach effective models, a combination of sources are used for training. These sources include information that is publicly available online and licensed information, as well as information from Meta's products and services," the giant said in a policy document.
"When we collect public information from the internet or license data from other providers to train our models, it may include personal information. For example, if we collect a public blog post it may include the author's name and contact information. When we do get personal information as part of this public and licensed data that we use to train our models, we don't specifically link this data to any Meta account."
- Meta lets Code Llama run riot under almost-open terms
- Meta trots out Llama 2 AI models, invites devs to hop on
- Of course Facebook will monetize an ad-generating AI
A spokesperson, however, told us that the biz did not train its latest large language model Llama 2 on any user data, "public or otherwise."
Eggheads at the Facebook titan have developed numerous generative AI models capable of producing text, images, code, and music. Meta top boss Mark Zuckerberg believes the technology is key to building his vision of the metaverse, where users can socialize and work in virtual reality worlds they can create. In the short-term, Meta is also developing generative AI tools for advertising and is using the technology to develop things like chatbots and customizable stickers for Facebook, Instagram, and Whatsapp in the near future. ®