Privacy Commissioner Rejects Meta’s Push to Weaken Australia’s Privacy Laws for AI

Meta argued that Australia is ‘out of step with international norms’ regarding privacy regulations.

A smartphone and a computer screen displaying the logos of the Instagram, Facebook, WhatsApp, and their parent company Meta in Toulouse, southwestern France, on Jan. 12, 2023. Lionel Bovaventure/AFP via Getty Images

Privacy Commissioner Carly Kind has rejected tech giant Meta’s proposal that Australia relax its privacy regulations to speed up development of AI.

Meta has recently made a submission (pdf) to an inquiry on digital technologies and the harvesting of personal data, where the company argued that Australia is “out of step with international norms” regarding privacy regulations.

This comes after the country has strengthened its privacy laws to enhance the protection of personal data, following reports that Meta used Australian Instagram and Facebook posts to train its AI in 2024.

During a parliamentary inquiry in September 2024, Meta representatives were questioned about why the company did not provide opt-out options for Australian users, as it did in the EU. However, the company was unable to provide an explanation.

“We are concerned that recent developments are moving Australia’s privacy regime to be out of step with international norms, impose obligations on industry that conflict with broader digital policy objectives to promote age-appropriate and safe experiences online, and disincentivise industry investment in AI in Australia or in pro-consumer outcomes,” Meta said in the submission.

The company also recommended that the Australian government adopt reforms that incorporate “innovation and economic interests as key objectives.”

Meta cited forecasts that AI could generate around US$53 billion (A$81.5 billion) to US$127 billion for the Australian economy by 2034.

At the same time, Meta highlighted the need for the company to use “real” user data for AI training.

“Meta has found that the nature of AI technology necessitates the processing of comprehensive real data to avoid biases and ensure language accuracy,” it said.

“For example, synthetic data is not workable because it needs to be generated by a generative AI model itself; therefore, in order for a model to even create synthetic data, it needs to first be trained on an appropriate data source.

“Generative AI models need to understand common concepts such as language, visual elements, and abstract ideas.”

A person reads Facebook posts on her phone in Manila, Philippines, on Nov. 21, 2024. Jam Sta Rosa/AFP via Getty Images

Privacy Commissioner’s Response

Kind rejected Meta’s argument that Australia’s current approach to privacy laws is not “pro-consumer.”

“It is hard to think of a more pro-consumer initiative than stronger regulation and enforcement of privacy protections, particularly in the digital domain,” she said on a social media post.

“The vast majority of Australians feel they lack control of their personal information, and want to see more regulation and fewer exploitative practices online.”

Kind cited a report by the Australian Competition and Consumer Commission, which indicated that 83 percent of Australians wanted tech companies to obtain user consent before using their data to train AI models.

Furthermore, Kind voiced her support for law reforms that focus on protecting users’ personal data.

“The privacy reforms proposed for Australia would be world-leading in their emphasis on ensuring that the handling of personal information is fair and reasonable, rather than enabling digital platforms to ‘consent their way out’ with lengthy terms and conditions that are impenetrable for most users,” she said.

“As the nation’s privacy regulator, it is my job to ensure that the application of the Privacy Act is in line with these community expectations.

“In my view, the role of the privacy regulator is to act in the public interest, not in the economic interests of tech companies.”

The privacy commissioner also noted that the guidance on developing and training generative AI models, which her agency published in October 202,4 does not affect Meta’s AI training, provided that the company follows proper processes.

“What will undermine the ability of actors like Meta to continue to develop AI models is declining public confidence that AI tools and systems are being designed to benefit society,” she said.

A Google researcher illustrates how AI is trainned in Aboriginal English in a project in Sydney, Australia, on Feb. 19 2025. AAP Image/Jennifer Dudley

Australia’s Privacy Law Reforms

In late November 2024, the federal parliament passed the Privacy and Other Legislation Amendment Act 2024, the first tranche of reforms aimed at strengthening Australia’s privacy laws.

A major change under the bill (pdf) is the introduction of a statutory tort of serious invasions of privacy, which effectively establishes privacy as a personal right.

This allows individuals to take legal action when their privacy is seriously infringed upon, with maximum compensation for non-economic loss and exemplary or punitive damages capped at $478,550.

The bill also criminalises the doxxing of personal data on the internet, with a maximum penalty of six years of imprisonment for single individuals and seven years for members of a group.

In addition, the legislation expands the power of the Australian Information Commissioner by allowing the agency to issue civil penalties, infringement and compliance notices.

Other changes include stricter data security requirements for organisations, and a new mechanism for overseas disclosure of personal information.

The second tranche of reforms is currently under development and expected to be introduced later this year.