Meta trains its AI on millions of Australians, but opt‑out now available
meta training
| Source: Mastodon | Original article
Meta has added a hidden option letting Australian users opt out of having their data used to train the company's AI.
Meta has quietly added an opt‑out mechanism that lets Australian users prevent their Facebook and Instagram content from being used to train the company’s generative‑AI models. The option, tucked inside the platform’s privacy settings, has been available since July 2025 but was never announced publicly, meaning most users are unlikely to discover it without guidance.
The move comes after sustained pressure from Australian regulators and privacy advocates who have warned that large‑scale data harvesting for AI training may breach the nation’s privacy laws. By offering a formal objection form, Meta is technically complying with the demand for user consent while keeping the feature hidden, a tactic that raises questions about transparency and trust.
For Meta, the ability to exclude Australian data could affect the breadth of material feeding its AI systems, potentially slowing model improvements that rely on diverse user‑generated content. At the same time, the hidden rollout may invite scrutiny from the Australian Competition and Consumer Commission and the Office of the Australian Information Commissioner, which have been monitoring the tech giant’s data practices.
What to watch next includes whether Meta will publicise the opt‑out more prominently, how many users actually submit objections, and if the company extends a similar mechanism to other jurisdictions facing comparable regulatory pressure. Observers will also be keen to see if Australian authorities demand a clearer disclosure or impose further obligations on Meta’s AI training pipeline. The episode underscores the growing tension between massive AI data needs and evolving privacy expectations worldwide.
Sources
Back to AIPULSEN