Australia's federal government will introduce draft Digital Duty of Care legislation this week that would require social media platforms to let users switch off their recommender algorithms. Platforms would have to ask users, through recurring pop-up messages, whether they want an algorithmically curated feed or content only from accounts they follow. Breaches would carry fines of more than A$100 million (R1.15bn).
Australian Communications Minister Anika Wells says many Australians will still choose the algorithm, but that big tech should offer the choice and respect it.
Interesting insights on Australia algorithm opt-out
The opt-out is one part of a wider bill. Platforms would carry a legal duty to identify and mitigate risks on their services, including preventing illegal content from being shared with adults. For under-18s, about six additional categories of psycho-social harm would be defined, covering areas such as body image and bullying.
The eSafety Commissioner and registered independent researchers could gain powers to audit platforms using accounts that pose as children, to see what gets served to minors. Australian Prime Minister Anthony Albanese says companion legislation will strengthen the Commissioner's evidence-gathering powers and double existing penalties.
This sits separately from Australia's under-16 account ban, which took effect in December 2025, and from a Greens bill by Senator Sarah Hanson-Young that has been before the Senate since April. Attorney-General Michelle Rowland confirmed on 1 September that the government is still weighing opt-in against opt-out, so the final model is not settled.
What others are saying about Australia algorithm opt-out
ABC News broke the details of the draft laws and the pop-up mechanism. The Conversation raises the definitional problem: Instagram runs separate recommender systems for its main feed, Stories and Reels, so opting out of the algorithm is not one action, and the expert view is that most users will not opt out at all. Startup Daily carries the sharper version of that critique, arguing a chronological feed by default with algorithmic content as the opt-in would change behaviour far more. Opposition leader Angus Taylor has warned against efforts to censor social media.
Regulating the product, not the user
Every major intervention so far has regulated access. Age bans, verification rules and parental controls all ask who may use the product. This asks how the product is allowed to work, and that is a different order of intervention. Recommender feeds are built on variable reward, the same intermittent reinforcement pattern that makes slot machines effective, and that design intent is not disputed by anyone who builds them.
What remains genuinely contested is the size of the resulting harm, particularly for adolescents, where researchers disagree sharply on how much of the observed decline in youth mental health can be attributed to these products.
Australia is legislating anyway, on the basis that the mechanism itself warrants a choice. For South Africa, the gap is obvious. Our draft age verification rules target video platforms and say nothing about recommender systems, and as with ChatGPT for Teens, we get whatever the platforms decide to ship globally.
You might also like our piece on the WhatsApp Business fees reshaping Meta's economics, how the Spotify AI persona label handles disclosure, and how SA online retail reached a tenth of all retail turnover.
Get more SA tech and business news and subscribe to The Open Letter.


