Meta admits to using Australian users’ Facebook and Instagram posts to train its AI

In Australia, user posts shared since 2007 have been trained by Meta’s AI, a group representative announced, without their consent being clearly requested. Only the data of users who opted for a private profile was spared.

22 million Australians on Facebook, 13 million on Instagram: a dream database for training artificial intelligence – but it would have been better to ask the main parties concerned for their consent. Melinda Claybaugh, director of privacy policies at Meta (Facebook, Instagram, WhatsApp, etc.), admitted that her company used the publications of its Australian users without asking their permission or allowing them to deactivate this data collection. She was heard on Tuesday, September 10 by Australian senators about the implementation of AI in the country.

“Is Meta using its Australian users’ posts shared from 2007 to the present to train its AIs?”asked a first senator, according to ABC. Melinda Claybaugh denies it. Another tries again, with more precision: “The truth is, unless a user has made their posts private since 2007, Meta uses every photo and message posted by its users, right?” «Correct»answers the director.

Posts by users under 18, however, would not have been used, the Meta employee defends herself. But when a senator asked her whether the photos he had posted on his own account of his child could be used by Meta, Melinda Claybaugh once again had to agree. Above all, the debate quickly turned to the possibility of allowing Australian users to validate or not the use of their data to train Meta’s AI – a question on which the representative did not pronounce herself, and which leaves little room for a positive answer in the immediate future.

European regulation, the benchmark for data protection

However, this is done in Europe, where Internet users can simply check or uncheck this option. An option which has nevertheless been the subject of a major standoff between Meta and the European Union, against the backdrop of the GDPR (general data protection regulation). “We have paused our AI products in Europe as long as there is ambiguity about the criteria for training artificial intelligence, Melinda Claybaugh justified herself, because of local regulation.”

In Australia, a user whose account is private is spared from data siphoning by Meta. An option that is currently considered sufficient for the firm to remain within the limits of local regulators, but which some politicians consider insufficient. Enough to push several Australian senators to question the toughening of national laws, which could be inspired by those known in Europe. “The personal moments that people share – photos, videos and recordings of people’s lives, their children and their families – are not resources that a technology company can turn into products.reacted Senator Tony Sheldon on X (ex-Twitter). This is an unprecedented violation of what makes us human. It is not only dishonest, it is predatory.”

And even in a framework as strict as the EU, Meta has multiplied attempts to test the limits of the authorities in terms of data protection, even if it means having to put its hand in its pocket. To its credit, the group has been fined, among other things: €405 million in September 2022 for failures in the processing of minors’ data at Instagram, €390 million for violation of the GDPR in January 2023, and even the gold medal in the field, reaching €1.2 billion in fines in May 2023 for having “continued to transfer personal data” of users from the European Economic Area (EEA) to the United States in May 2023.



Source: www.liberation.fr