Meta using Instagram and Facebook posts from as far back as 2007 to train artificial intelligence tools
By Lina ElsaadiIf you're among the majority of Australians with Facebook or Instagram accounts, your social activity on those platforms is likely being used to train Meta's artificial intelligence (AI), and there's little you can do to stop it.
There has been growing concern after Meta – the parent company of Facebook and Instagram – flagged a new privacy policy that was to take effect in Europe and the UK from June 26.
Under the policy – which Meta has now agreed to pause while it answers further requests from regulators – it had planned to use data from the public posts of Facebook and Instagram users to train and improve its AI tools.
The pause is at the request of the Irish Data Protection Commission on behalf of European Data Protection Authorities, and Meta has expressed disappointment, saying it will delay bringing the benefits of AI to people in Europe.
In Australia, Meta is actually already using public users' data in such a way, according to its privacy policies.
As it stands, the European Union (EU) is one of the only jurisdictions in which users can opt out of having their Meta accounts used to train AI.
This is because they have strong privacy protections, in the form of the General Data Protection Regulation (GDPR).
Australia's protections are weaker, which is why Meta can use Australian users' data without giving them a chance to opt out.
So, how do Australians feel about that?
When asked on the street, most Australians responded with "I did not know that" – and they weren't happy.
There were plenty of replies like "That's kind of concerning", as well as "Feels a bit like an invasion of privacy" and "Oh, that's not good. That feels bad. I don't like that".
There were also handfuls of "I'm not that surprised".
Users give Meta permission to use their data this way when they agree to the terms of service on the platforms.
But when asked, most Australians admit they've never read them.
It is worth noting that Meta's approach is not too different from that taken by other leading AI labs.
Meta said that to properly serve users in different countries, models that powered AI needed to be trained with relevant information.
It does not use private messages or data from accounts of those aged under 18.
Meta told the ABC in a statement, "We're committed to building AI responsibly and believe it's important that people understand how we train the models that power our generative AI products.
"With the release of our AI experiences, we've shared details about the kinds of information we use to build and improve AI experiences – which includes public posts from Instagram and Facebook – consistent with our privacy policy and terms of service.
"Since no AI model is perfect, feedback is instrumental to AI's continued development, and we're engaging with experts, policy makers, and advocates as we strive to build AI responsibly."
'I have a skill that I've worked on, and it's mine'
Children's Illustrator Sara Fandrey lives in Portugal. When she found out about the proposed changes in the EU, she was furious.
"We were like, we need to tell as many people as we can," she said.
She posted a video to Instagram, showing how to fill out an objection form – which she could access, as Portugal is in the EU.
That video now has over 7 million views.
"I had absolutely no idea it would become this viral, and would start a series of things, [like] people closing down their accounts."
Videos like Sara's began cropping up on all kinds of social media platforms, and the hashtag #noaiart was born.
Searches for "Instagram ai opt-out" on Google increased, and with it, articles explaining how to opt out.
"I was completely shocked," Sara said.
"How can a multimillionaire company like Meta take what's mine, what's my intellectual property, and just scrape it off and feed it to their AI?"
Meta's AI Creative Tools can already generate images for you.
And that's another sore spot for artists like Sara.
"These images do not come from thin air. They're all based on other images. I matter. I am a human being. I have a voice. I have a skill that I have gained that I've worked on and it's mine."
As an artist, she is concerned that her livelihood is going to be impacted by this. She's not alone in that opinion.
Thomas Fitzpatrick is a Sydney-based Illustrator.
Since he can't opt out of Meta using his data in this way, he is considering other options.
"I'll probably stop using Instagram. I have no interest in having my work trained in algorithms and models, so I'll honestly probably just get rid of it."
He's concerned for the future of the industry.
"For illustrators and creatives who are starting out, it's going to take a lot of work away from them," he said.
"If you're a small business, it's very cheap to just type something in and get results straight away. So why would you pay an up-and-coming illustrator?"
How is Meta allowed to do this anyway?
Dr Joanne Gray, a lecturer in digital cultures at the University of Sydney, explains: "The precedent in the US suggests that these companies are doing it under fair use, a US exemption that allows you to do some copying and create something new.
"These models copy style. While it's not copyright infringement to copy style, it is important as an economic asset to people when they become established.
"And so, it's a threat to their business in a new way, outside of copyright."
This is all so new and unprecedented, she explains, that "we're really yet to see how that's going to play out in terms of legal cases, but also in terms of how creators might be remunerated."
Dr Gray adds: "We need to do more to ensure that we have better control over our data and that these corporations don't have unfettered access to make money and trade on data that's created by us and about us."
Is this the new normal?
Meta seemingly doesn't plan to get rid of AI anytime soon.
They've now added a "Made by AI" label, requiring users to have all realistic appearing AI-generated content carry it.
Artists have taken note and are migrating to apps like Cara – an artist-run, anti-AI social platform.
Its user base has recently grown from 40,000 to over 650,000.
Cara incorporates a project called Glaze, "a system designed to protect human artists by disrupting style mimicry in the training of generative AI models".
As major players like Meta make moves in the AI space, users will continue to react accordingly.
Editor's note 25/6/2024): This story originally incorrectly stated that Australian users would be subject to new privacy policies by Meta that were planned to come into effect in the EU/UK on June 26. The story incorrectly stated that both public and private account data would be used. Australian users will not be affected by the planned EU changes, and public Australian data is already being used by Meta to train AI, according to its privacy policies.