Credit: via Canva
Australia has published draft legislation that would require social media platforms to show users recurring prompts offering to turn the algorithmic feed off.
The exposure draft of the Digital Duty of Care bill went out this week, Reuters reported, and carries penalties above A$100m for breaches.
Plenty of jurisdictions have established that users should be able to escape recommender systems; almost none require the platform to keep asking whether they would like to.
A right nobody knows about, buried three levels into a settings menu, is worth roughly what the platform wants it to be worth, and this bill is an attempt to price that in.
The wider duty of care runs alongside it. Platforms would have to identify and mitigate risks on their services, prevent the sharing of harmful and illegal material, and take specific measures to protect minors from content relating to body image, bullying and other psycho-social harms.
The eSafety Commissioner and independent researchers would gain the power to monitor platforms using undercover accounts posing as children, which is a research method the companies have generally resisted.
Europe reached the first half of this in 2023. The Digital Services Act requires the largest platforms to offer at least one recommender option not based on profiling, which is the same substantive right Australia is legislating.
What the DSA does not do is oblige anyone to surface it, and take-up of the chronological feeds now available on the major services has been correspondingly modest.
Australia’s draft would put the choice in front of the user on a repeating basis, which is a materially different instrument aimed at the same problem.
The politics are not settled. The Coalition has arrived at two positions at once: shadow communications minister Sarah Henderson has attacked the “insidious psychological manipulation” of children by recommender systems, while opposition leader Angus Taylor said he was “deeply sceptical” and feared the bill was an attempt by the government to censor social media.
Shadow treasurer Tim Wilson has raised similar concerns about restricting opinion. The Greens, from the other direction, want algorithmic feeds to be opt-in rather than opt-out.
That disagreement is the substantive one. An opt-out regime leaves the default in place and relies on users acting; an opt-in regime would switch personalised ranking off until someone asks for it, which would be the most aggressive intervention any democracy has made in how feeds work. The bill as drafted takes the softer path and compensates with the prompt.
None of this arrives in a vacuum. Australia has spent two years building the most interventionist online-safety regime in the democratic world, and the results have been uneven.
Its under-16 social media ban works better on paper than in practice; the regulator has found Meta, TikTok and YouTube not complying with it, and the government has moved to double the fines and expand the regulator’s powers in response. Age checks failed at the first test.
This is the reason to watch the enforcement provisions rather than the headline duty. A pop-up requirement is trivially auditable in a way that age assurance is not: either the prompt appears at the stated interval, or it does not, and an undercover researcher account can establish which within an afternoon.
For a regime that has struggled to verify its own rules, that may be the more consequential design choice.
The exposure draft is a consultation document, not a law. It goes to parliament after that process, into a chamber where the opposition is split, and the crossbench wants something stronger.
Get the TNW newsletter
Get the most important tech news in your inbox each week.
