Online Safety Amendment (Strengthening Enforcement for the Social Media Minimum Age) Bill 2026

Senator HENDERSON (Victoria) (19:13): I just want to make some quick remarks. I did speak earlier about the algorithmic transparency requirements and the amendment that we have brought forward, and I am looking forward to the minister's answer to Senator Hanson-Young's question, but I just want to outline why this is so important. The starting point in understanding any harm that the algorithms of social media companies drive is, 'What is going on?' As Senator Hanson-Young has correctly said, what is going on under the hood?

What is the business model? That is why this amendment—and obviously this can be seen in Hansard, but this is for those listening to this debate—in relation to the algorithm seeks: (a) for each system used materially to select, rank, recommend, promote, demote or otherwise determine the prevalence or prominence of content, accounts or search results for end-users in Australia … (i) the algorithmic logic and decision-making architecture of the system; and (ii) the material factors, inputs and signals relied upon; and (iii) the material parameters, thresholds and weightings applied; and (iv) an explanation of how those factors affect the content presented to end-users … In effect, that is the content recommendation algorithm.

Then this amendment also requires the social media platforms—and these are the large social media platforms with turnovers of more than $100 million—to unlock their content moderation algorithm. This is very important. The government's talked a lot about transparency and about protecting children online.

We want to see how these social media platforms are moderating content and how this information is being used to identify, classify, remove, restrict, promote, demote or otherwise moderate content or accounts. Again, we want to see the algorithmic logic, the material rules and parameters and the like, the circumstances in which content or an account is referred for human moderation to make sure that it's not the bots at work and that there are humans actually involved, the policies and rules governing automated and human moderation and any information about the accuracy, error rates and material limitations.

The other key part about this algorithmic transparency is understanding the age-assurance algorithm. Again, we are seeking, in this amendment, the algorithmic logic, the manner in which the age-assurance algorithm is implemented, the categories of information and data relied upon, the material decision rules and the circumstances in which an end user is referred to another method of age assurance.

These are hugely important issues in relation to the enforcement of the social media ban. We keep hearing about the age assurance that's been put in place, and yet we know it's not working. Then also, very importantly, this amendment is not seeking proprietary information, such as the source code.

There are very important exemptions. For protection, safety, security and privacy, personal information must not be disclosed, nor material which compromises the safety or security of an end user or the platform, nor material which may assist a person to circumvent or defeat systems designed to protect children or prevent cyberattacks, fraud, spam, platform manipulation or other malicious activity.

What's really concerning in relation to the government's position, Minister, is that there is not any of this in the digital duty-of-care exposure draft. How can the eSafety Commissioner determine whether a social media platform is taking reasonable steps to keep children off their platform if we're not able to understand what's going on behind closed doors, how these algorithms are being used to drive harm or not properly moderate content or, in the case of children under the age of 16, how the age-assurance algorithm is actually working?

It is concerning because it indicates, firstly, that either the government is not listening to the advice it is receiving, including in our Senate inquiry, or the government doesn't understand why algorithmic transparency is critical. So I ask you, Minister: why have you taken the view that you should not support this particular amendment, and why is this not in the digital duty-of-care exposure draft?

What this indicates is that, if you're not prepared to hold big tech to account and ask them for information about their algorithms, you're not prepared to do the job. I say 'you' as in the government. We have certainly seen, on X's For You feed, that X has published all the algorithmic parameters which give users a very good insight into how decisions are being made as to how content is being served.

So could you please provide some insight into why the government is turning its back on this very important amendment.