Meta, the parent company of Facebook and Instagram, has confirmed that it has been utilizing publicly shared posts and photos from adult users on its platforms to train its artificial intelligence (AI) models since 2007. This admission came during an Australian government inquiry, where Meta’s global privacy director, Melinda Claybaugh, initially denied the practice before ultimately acknowledging it.
Initial Denials and Confirmation
During the inquiry, Melinda Claybaugh was questioned about whether Meta had used Australian users’ data for AI training. At first, she denied that data from as far back as 2007 was used, telling Labor senator Tony Sheldon, “We have not done that.” However, Greens senator David Shoebridge challenged this claim, pointing out that unless posts were set to private, Meta had indeed been scraping public data from Facebook and Instagram for years. Confronted with this evidence, Claybaugh admitted, “Correct,” confirming that public posts, including those from children, were collected.
Lack of Opt-Out Option for Australians
The inquiry also revealed that Australian users do not have the option to opt out of this data collection, unlike users in the European Union (EU). Claybaugh acknowledged that while EU users can refuse consent for their data to be used in AI training, Australians do not have a similar choice. This discrepancy is partly due to stricter privacy laws in the EU, which have prompted Meta to offer an opt-out option there.
When questioned, Claybaugh confirmed that even publicly shared photos of children, if posted by their parents, would be used for AI training. She also could not clarify whether data from users who were minors at the time of their posts but are now adults had been collected.
Meta’s Data Collection Policies
Meta’s privacy policy states that the company uses public posts and comments on its platforms to train its AI models. However, it is unclear when this data collection began and how extensive it has been. In response to inquiries, Meta has indicated that only posts set to “public” are scraped, and while future posts can be protected by changing privacy settings, past data remains in Meta’s systems.
European vs. Australian Data Policies
One of the key takeaways from the inquiry was the contrast between Meta’s data handling practices in Australia and the EU. While Australians are not given an opt-out option, EU users can choose to refuse data collection for AI purposes. This difference is attributed to the EU’s stringent privacy regulations, which have led Meta to pause AI product launches in Europe until legal uncertainties are resolved. Claybaugh explained that the EU’s regulatory environment has necessitated these precautionary measures.
Privacy Concerns and Regulatory Responses
The inquiry has intensified concerns about Meta’s data practices, particularly in light of the lack of transparency and the inability of users to control how their historical data is used. The issue is compounded by the fact that users who posted content as far back as 2007, often without knowledge of its future use, have limited options for recourse.
As the Australian government considers stricter regulations on social media, including a potential ban on platforms for children, Meta’s data practices are under increasing scrutiny. The company’s approach to data collection and privacy is likely to be a significant focus of ongoing debates about how to balance innovation with user rights.
