Statistical models are not technically algorithms; they are a heuristic which "approximates" a solution probabilistically. They use algorithms to function and operate on data (like everything a computer does) but they are not algorithms themselves.

The main thing they have to do is clearly define that results derived by statistical modeling are opt-in, set limits on what personal data can be used to present content, and ensure there are significant penalties for crippling the experience of everyone who hasn't opted-in, which all are guaranteed to do to force users to opt-in.

They should define a bare minimum baseline of data points that should be made available to users to control the content they want to see using non-statistical algorithms (e.g. time, location, content type or popularity, friends/subs/followed only or ones they recommend themselves, etc). They should also define that opt-in/out (enabling and disabling statistical model results) should be a part of the main filter UI that uses the aforementioned data points; a simple toggle at the top of their page/feed, permanently visible and near-instantly applicable; otherwise they're guaranteed to hide it away and make opt-in/out a major friction point.

Basically you need to treat them like they're sociopathic criminals who will never act in good faith voluntarily, because all multinationals and large corporations are lead by criminally corrupt sociopaths.

The wording in the article body seems reasonable

Users can choose to opt in to having their default feed include personalised content recommended by the algorithm. Alternatively, users can opt out of personally recommended content in their default feed and see the friends and creators they choose to follow.

But then the actual quotes on the bottom don't use this wording. I'm not sure what to expect.

Knowing the Australian government and their long track record of not understanding technology I think it's a given that it will be phrased badly and be unworkable.