Meta executives confirm tech giant scraping public data since 2007

Tech giant Meta has been systematically scraping the data of millions of Australian Facebook and Instagram users since 2007 to train its AI products, a senate hearing into the company has revealed.

Meta executive Melinda Claybaugh, appearing before the senate select committee on adopting artificial intelligence on Wednesday morning, confirmed the massive data collection policy under questioning from Labor senator Tony Sheldon.

“We are using public data to train our foundation model and the services we build on that model,” she said.

“We are using publicly, widely accessible data that is available online.”

Ms Claybaugh initially denied the company pulled photos and text from Facebook or Instagram profiles but then clarified that if a user had set their posts to public, which means the photos or text are accessible to anyone, then the company collected the information and fed them into its AI training models.

“We are using public data from our product and services,” she said.

She said unless a user had consciously set posts to private, the company had scraped the data.

Ms Claybaugh said the company did not collect data from profiles under the age of 18.

However, if a parent had posted a photo of their child, that photo could be scraped, she said.

Greens senator David Shoebridge said the data collection method posted “ethical issues” and argued an Australian mother posting a photo of her child to celebrate a milestone would likely not have contemplated the photo being used to train AI.

“That mum didn’t give Meta permission to learn from her child,” he said.

Ms Claybaugh responded that the company had a “layered approach to the mitigations we have in place”.

“Personal data cannot be spit out of the generative AI product, cannot be associated with a particular person,” she said.

Ms Claybaugh appeared alongside Meta vice president public policy for APAC Simon Milner.

The executives fronted the committee as the parliament considers a ban on children accessing social media.