Australia (Commonwealth)_Meta has been utilizing public Facebook and Instagram posts from Australian users to train its artificial intelligence systems, contrary to recent media reports suggesting the data usage would commence in late June. The social media giant had planned to start leveraging public posts from users in Europe and the UK for AI training on June 26. However, this move has been postponed, leading to confusion among users in other regions regarding their data’s involvement.
Despite the postponement in Europe, Meta has already been using public posts from platforms like Facebook and Instagram to train its AI models in countries such as Australia and the United States, where data privacy regulations are less stringent compared to Europe. The Office of the Australian Information Commissioner (OAIC) has requested a meeting with Meta to discuss the company’s AI products and the associated privacy safeguards.
A spokesperson for Meta did not deny the claims about using public posts from Australian users for AI training. They stated that Meta had shared details about the types of information used to develop and enhance AI experiences, which includes public posts from Instagram and Facebook, in accordance with the company’s privacy policy and terms of service. The spokesperson emphasized Meta’s commitment to building AI responsibly and bringing innovative solutions to people and businesses globally, noting that using publicly available information for AI training is a common industry practice.
Meta clarified that it does not utilize posts from private accounts or from users under the age of 18 for training AI models. In a blog post from September 2023, Mike Clark, Meta’s director of product management for generative AI data and safety, indicated that the company had started using publicly shared posts from Instagram and Facebook, including photos and text, to aid in training its generative AI models. Clark also mentioned that messages sent to Meta’s AI chatbots were included in the training data, but private messages on Messenger or WhatsApp were not used. Instead, Meta relies on publicly available information online, licensed data, and information from its products and services.
The OAIC encourages Australians to regularly review and update their privacy settings to avoid unintentionally sharing information they prefer to keep private, thereby preventing their posts from being used to train AI models. Under the Australian Privacy Act, privacy obligations apply when personal information is used to train AI systems. Organizations can only use such information for purposes beyond the original intent if the individual has consented or if it is reasonably expected.
Unlike users in Europe and the UK, Australians currently do not have the option to opt out of their public posts being used for AI training by Meta. This discrepancy arises because Europe’s General Data Protection Regulation (GDPR) imposes stricter data collection regulations. Australians, however, can submit requests regarding the use of their personal information by third parties for AI training and delete information provided to Meta’s AI chatbots to correct inaccuracies.
The Australian government has yet to fully implement the recommendations from the Attorney-General’s review of the nation’s Privacy Act released in February 2023. Recently, the Australian eSafety Commissioner signed an administrative agreement with the European Commission to support each other in regulating online platforms, including areas involving algorithms and artificial intelligence.
In Europe, Meta had notified users about changes to its privacy policy, scheduled for June 26, which would allow the use of public posts for AI training. This announcement faced backlash from European users and privacy groups like Noyb, which filed complaints urging data privacy authorities to take action. Following a request from the Irish Data Protection Commission on behalf of European authorities, Meta paused its plans for Europe. Meta’s global engagement director of privacy policy, Stefano Fratta, expressed disappointment, claiming it was a setback for European AI innovation and competition.
Despite the pause, Meta remains confident that its approach complies with European laws and regulations. Noyb chair Max Schrems criticized Meta’s approach, suggesting that the company could obtain valid consent for AI training but chooses not to, framing Meta’s actions as a form of “collective punishment” for the entire continent.
