Meta confesses, but there’s no opting out for Aussies
Facebook parent company Meta has admitted to scraping posts and photos from Australian users since as far back as 2007 to train the company’s AI models. This confession came about at an inquiry that was held on September 10 in Australia, when Meta’s global privacy director, Melinda Claybaugh, was asked about the company’s efforts in the AI space – particularly on matters relating to the privacy of users.
Per the ABC, Claybaugh confirmed that posts and photos were taken from Aussie users that were registered as over 18 years old to feed the company’s large language model (LLM) development, although anyone registered under that age limit was exempt. However, photos of children uploaded on other accounts, such as by their parents or other adult family members, were fair game and could have been scraped.
The most glaring admission made during the inquiry was that there is no opt-out option for Australian users unless they set all their posts to private. Such an option exists for users in the European Union where strong privacy and data protection laws are enforced.Explaining Meta’s AI content scraping
Meta using Facebook and Instagram posts to train its AI algorithms is not a revelation. It’s an international practice for the tech giant and was revealed in a blog post last September. At the time, the tech giant made it clear that “We didn’t train these models using people’s private posts” and that “We also do not use the content of your private messages with friends and family to train our AIs”.
Even if posts were all set to private by an Australian user after the fact – for example, publicly available posts from 2010 being made private on hearing this news – there remains the question of whether that information would stay within the LLM’s database.
