An Australian Senate committee has questioned tech giant Meta about the ethics of its decision to use photos of individuals to train artificial intelligence (AI) without user knowledge.
Labor Senator Tony Sheldon, chairman of the Select Committee on Adopting AI, asked Meta’s representatives about whether the company used Australian Instagram and Facebook posts as far back as 2007 to train its AI model, as reported by the media in June 2024.
Meta Global Privacy Policy Director Melinda Claybaugh responded that her company was using “public data” from its products and services.
“That means when you go on Facebook or Instagram, and you make a post, you select the audience for that post, that statement, that photo, whatever it is you’re posting online,” she said.
“If you choose to make that post–the text or the image–public, that is publicly sharing that information.”
Greens Senator David Shoebridge pressed her further, demanding clarification.
“Since 2007, Meta has just decided you will scrape all of the photos and all of the text from every public post on Instagram or Facebook that Australians have shared since 2007 unless there was a conscious decision to set them on private. That’s actually the reality, isn’t it?” he asked.
“Correct,” Claybaugh admitted.
At the same time, the director dismissed concerns that Meta was using data from adolescents’ accounts to train its AI.
“We are using public photos posted by people over 18,” she said.
Meta Questioned About the Ethics of Its AI Training
Shoebridge questioned Meta about the ethics of its AI training, saying that some users did not give the company permission to use their images.“Don’t you see there’s an ethical problem here from Meta doing this?” he asked
In response, Claybaugh said Meta had taken measures to mitigate problems with using its users’ personal data.
“We take a lot of steps, both in collecting the data, training the model, applying filters and outputs to make sure that personal data cannot be spit out of the generative AI product, [and] cannot be associated with a particular person,” she said.
“I can assure you that our AI development process is an end-to-end accountability process where we are taking mitigations at every step of the development …, and particularly testing for risks around privacy safety.
No Opt-Out Option for Australian Users
Sheldon pointed out that Meta provided an option for EU users to opt out of having their data used to train AI and asked why the company did not do the same in Australia.Claybaugh said this was related to EU’s privacy law.
“In Europe, there is an ongoing legal question around the interpretation of the existing privacy law with respect to AI training, so we have paused launching our AI products in Europe while there is a lack of certainty,” she said.
“You’re correct that we are offering an opt-out to users in Europe. However, that is not a settled legal situation, and I will say that the ongoing conversation in Europe is the direct result of the existing regulations.”
When pressed about whether Meta would provide such an option to Australian users, Claybaugh did not give a direct answer.
“Is it happening now?” Sheldon asked.
“It is not happening now,” Claybaugh replied.
Australian Children’s Personal Data Misused for Training AI
Claybaugh’s testimony comes amid rising concerns about the misuse of children’s personal photos in AI training.The human rights organisation said this could pose privacy risks as malicious actors could use those images to generate explicit imagery of children.
Hye Jung Han, a researcher and advocate at the HRW, urged the Australian government to introduce laws to protect children’s data from the misuse of AI technology.
“Children should not have to live in fear that their photos might be stolen and weaponised against them,” she said.
