Your Facebook and Instagram posts are a rich resource for ‘training’ artificial intelligence – and Meta, the parent company of Facebook and Instagram, is using data from both services to ‘train’ its artificial intelligence (AI).
It became clear earlier this year that Meta is ‘training’ its AI models using photos, posts and comments from users, after users in Europe were sent a message about an update to the privacy policy.
Meta’s privacy policy says: “Since it takes such a large amount of data to teach effective models, a combination of sources are used for training.”
The company says that publicly available information from the internet is used to train models, along with what the company describes as “information shared on Meta’s Products and services. This information could be things like posts or photos and their captions. We do not use the content of your private messages with friends and family to train our AIs.”
But there is a way to request to opt out from having your data used in this way.
In the training phase, AI models are ‘fed’ curated data – in this case, your Facebook and Instagram photos and posts – to make them respond more accurately.
In the inference phase, users ask a query – for instance, you might ask a large language model such as ChatGPT to write a list of salad recipes – and the model will draw on its store of information that it has been trained on in order to answer the question.
Rather than ‘storing’ the full data, the model stores data as ‘weights’ which relate different facts, images and ideas, which are then used to deal with fresh commands.
People worry about private data being used in this way, as the data is out of their control and it’s not always clear how AI models use it – or whether there’s a risk of personal data resurfacing.
Even if you opt out of Facebook using your data, models may be being trained on your posts from other social media platforms.
Tech companies including Meta, OpenAI and Alphabet’s have been criticised for using information scraped from the internet without permission to train their AI models.
These companies are weighing how to handle the private or copyrighted materials vacuumed up in that process that their AI systems may reproduce, while facing lawsuits from authors accusing them of infringing copyrights.
Facebook offers a dedicated page where users can object to their data being used to train AI, although it stops short of promising that it won’t.
Users can object by sharing their email address, their country of residence, and why they believe their messages and images should not be used to train AI. The form can be found here.
Meta say: “We’ll review objection requests in accordance with relevant data protection laws. If your request is honoured, it will be applied going forward.”
Speaking to Yahoo News, a Meta representative explained that any reasonable request not to have data used will be honoured (and users don’t have to prove that they have been harmed or are a special case).
But Meta says that your data may still be processed if you are (for example) featured in an image posted by someone else.
The company says: “We may still process information about you to develop and improve AI at Meta, even if you object or don’t use our products and services. For example, this could happen if you or your information: appear anywhere in an image shared on our products or services by someone who uses them, are mentioned in posts or captions that someone else shares on our products and services.”