Meta has quietly introduced a way for Australians to stop the tech giant from training its AI on their user data, but the company never announced it, leaving the choice effectively hidden for more than a year.
There was furious public backlash in mid-2024 when it was revealed Meta was using photos, posts and comments stretching back to 2007 to train its AI.
At the time, Australians were told that unlike EU citizens, they could not opt out because of weaker privacy protections in Australia.
But ABC News has learned an “objection” form has been available locally since July 2025, buried within privacy settings.
“You have to go digging through their privacy policies and their support pages … certainly they haven’t been doing outreach to users,” said Tom Sulston from Digital Rights Watch.
“A lot of people who would like to use that option are not going to find out about it.
“And you can understand why — there’s a commercial imperative for [Meta] to get their hands on as much of our data as they can.”
The objection form has been available in the EU since 2024, but the ABC has confirmed that no separate announcement was made when it was added in Australia a year later.
Nor have there been any reports of users being notified directly by Meta.
Instead, the information has only recently been filtering out via user-generated posts.
“While it’s good that we can opt out … we haven’t really been given a choice if we’re not made aware of that choice in the first place,” Mr Sulston said.
How to opt out of Meta’s AI training
A step by step guide:
Alternatively, if you’re logged in to your Facebook account, the form can be found here.
According to Meta, users who submit a form will get an email confirming the company will not use their public data to train its AI in the future.
It also explains that accounts set to “private”, are already exempt, as are all private messages, and accounts listed as belonging to under 18s.
Why did Meta start offering Australians an opt out?
Meta has not shared why it made the decision and the relevant sections of the Australian Privacy Act have not changed in recent years.
One possible explanation is that the question of how the law applies is not entirely settled.
The Office of the Australian Information Commissioner (OAIC) previously issued public guidance that “just because data is publicly available … does not mean it can legally be used to train … AI models”.
Since Meta’s AI training practices were first revealed, the OAIC has been assessing the company’s behaviour.
“The issue of whether social media platforms are permitted under Australian privacy law to train AI models on user data is a matter the OAIC has had under consideration for some time,” a spokesperson said.
According to Meta, users who submit a form will get an email confirming the company will not use their public data to train its AI. (REUTERS/Yves Herman)
“We have engaged extensively with Meta and other platforms, and intend to publish the results of our investigations in the coming months.”
Some privacy advocates believe Meta’s decision might be a response to that legal ambiguity, or even a pre-emptive move, as the government mulls further changes to privacy rules.
“By doing something, [Meta] can claim that they’ve done something, and then it’s easier for them to use our data everywhere else to train AI,” Mr Sulston said.
“The problem is, of course … we’re not informed about how we might make that decision.
“So it lets them meet some obligations … but at the same time it doesn’t really fundamentally change anything”.
Meta was approached for comment.
Source: www.abc.net.au
