Meta: Have You Been Posting Actively on Facebook? Your Posts Since 2007 Have Trained Its AI.
Meta’s Use of Public Data for AI Training
Overview of Data Usage
Since 2007, Meta (formerly Facebook) has utilized a vast array of publicly shared information from its platforms, including Facebook and Instagram, to train its artificial intelligence (AI) models. This includes not only text but also images posted by users who have not set their accounts or posts to private. Melinda Claybaugh, the global privacy director at Meta, initially denied claims during a government investigation that user data was utilized for AI training, but later clarified that Meta did collect this data.
Public Scraping Practices
Senator David Shoebridge, a member of the Green Party, highlighted that Meta has essentially made a decision to scrape all public content, which includes photos and text from every public post on Instagram and Facebook since 2007. This extensive collection continues unless users specifically opt to make their posts private.
Meta’s Official Position
According to information from Meta’s privacy center, the company maintains that it gathers publicly available content from Facebook and Instagram for the purpose of training its generative AI. They assert that this process helps enhance features on their platforms and benefits the open-source community. However, there is little clarity on how long Meta has been collecting this data or when these practices began.
In a response to a query from The New York Times, Meta confirmed that only content set to public prior to a change in privacy settings would be available for scraping in the future. It is important to note that any previously collected data will remain stored, raising concerns about users who shared content during their youth, unaware that such information could be utilized commercially.
Concerns Over Minor’s Data
Claybaugh emphasized that Meta does not collect data from individuals under eighteen, although concerns remain about posts made by users who were minors at the time. Senator Tony Sheldon raised questions about whether Meta would remove publicly accessible images of his children, to which Claybaugh confirmed they would comply. However, it was unclear if the same protocol would apply to adult accounts created while the users were still minors.
Regional Privacy Protections
In different regions, privacy regulations vary widely. For instance, users in Europe have the opportunity to opt out of allowing their data to be used for AI training, which was recently prohibited in Brazil. Unfortunately, users elsewhere, including Australia, do not have the same privilege. The absence of robust privacy laws allows Meta to continue utilizing public posts without users’ explicit consent.
Shoebridge indicated that if Australia had similar protections in place, local data would have remained safeguarded. This disparity raises concerns about how children’s images and videos are utilized and exploited on platforms like Facebook and Instagram, particularly due to the lack of action from governments on privacy issues.
Frequently Asked Questions
Has Meta trained its AI on my posts from Facebook and Instagram?
Yes, Meta has trained its AI models on publicly visible posts and images from Facebook and Instagram that were shared since 2007, as long as users did not explicitly set them to private.
Is it possible to prevent Meta from using my data for AI training?
You can prevent Meta from collecting your future posts by setting them to private. However, any public posts made before this change will still be retained by the company.
As the conversation around data privacy progresses, users are encouraged to be mindful of their sharing habits and privacy settings on social platforms, considering the implications of their digital footprints.
