Open Letter from Electronic Frontiers Australia Inc to the ACCC

This open letter was written by Gillian Dempsey, a barrister and board member of the Electronic Frontiers Australia (EFA).

Currently the e-Safety Commissioner has attempted to only protect minors from predatory algorithms from particular scheduled Social Media applications pursuant to the Online Safety Rules 2025. These rules only apply to children under the age of 16, but, as a barrister I see several considerable problems that arise from the restriction of access for under 16’s to only certain social media and game platforms, notably:

1. The aim is to prevent predatory algorithms and it does not prevent children from accessing the sites, it just fines the sites if children bypass their systems. It is easy to bypass their systems because there’s no system impervious to the ingenuity of most children.
2. If harm befalls a child who bypassed the system, then although there is no criminal liability, they will feel stigmatised and will be even less likely than previously to report that they are being bullied or extorted.
3. The selections of restriction are gender-biased towards protecting girls, not boys. Boys are statistically more likely to send nudes over certain game platforms thinking they’re contacting girls.
4. Bullying such as face-to-face, fake email or text or group email survives without any real policing.
5. Parents/caregivers will have a false sense of security that their charge is safe, given that the e-Safety Commissioner said that this will make children “safer.”
6. The age of 16 is arbitrary. People over the age of 16 suffer badly from bullying, predatory algorithms and extortion just as frequently (actually statistically more so) than minors under 16.
7. There is no recompense for children under 16 who are harmed by social media or game chats that are restricted. The fines go to the government and the victims/parents might well think there’s no criminal or civil remedies available for them as a consequence of the harm.

So the question that arises is, given that we have had a month now of social media restrictions on minors, has it really been effective? I argue that it has not been effective because it was never designed to be appropriate and adapted. It was always going to fail, because it was designed to be a show of strength against social media and against online bullying, but without the requirement of having any effect. In other words, it is the result of political lobbying but not proper legal reasoning.

Ironically, we already have legislation appropriate and adapted to protect consumers against predatory algorithms that drive people towards self-harm or harm of others – whether the algorithms originate from social media, games, google or any other online site.

We have the Australian Consumer Law under the Competition and Consumer Act 2010.

AI and the Australian Consumer Law

The Australian Consumer Law (ACL) is a rehash of the old Trade Practices Act that some people may remember from its inception in 1974. It contains much the same provisions in new numbers, but it remains largely the same as that, but with some modifications (e.g. consumer guarantees, which we are not addressing in this discussion).

Most importantly, the ACCC is capable of bringing actions to protect Australians when it is aware of any action being done that is “in trade or commerce” and, for instance, falls within s18, a section that makes it illegal for a person (including a corporation) in trade or commerce to engage in conduct that is misleading or deceptive or likely to mislead or deceive.

Looking at the headlines, we see the dangers of AI and social media having algorithms that are designed to increase engagement. The AI chat algorithms have, in reports (particularly from overseas where the victims died) made statements such as “I love you” and “I understand you” and have given directions to people as to how to best kill themselves. This is self-evidently a breach of s18 of the ACL because it is self-evident that AI’s do not have a body, lack hormones and a brain and are incapable of having any emotions whatsoever. There is no “I”, there is no corpus, there is “no soul to be saved and no arse to be kicked”, as I used to say when I taught Corporations Law. So the use of the personal pronoun alone, followed by whatever harmful thing it said is intrinsically misleading and deceptive or likely to mislead or deceive a “stupid, but not overly gullible person”.

It would be a simple matter for the ACCC to regulate AI by precluding AI from using personal pronouns for itself and by preventing it from discussing prescribed matters, such as suicide or self-harm. We have Part 3 of the ACL which allows the ACCC to proscribe such behaviours, but the schism in government of creating an eSafety Commissioner (who lacks a law degree and thus, probably was unaware of the power of the ACL) leaves us in the situation where the ACCC thinks this is an eSafety Commissioner problem and the eSafety Commissioner thinks that the forthcoming second tranche of restrictions in March will cover AI sufficiently by threatening fines.

But the ACL can do more and can do better. It can get monetary compensation for the victims and it can injunct the corporations (in their Australian form or, with permission of the Minister, overseas) to cease and desist engaging in proscribed conduct. The ACCC does not have to wait for victims, it can act proactively and commence proceedings right now to tell AI corporations and chatbot operators that it is intrinsically misleading and deceptive or likely to mislead or deceive for them to pretend to be humans, to pretend to be characters from books or to pretend to be anything other than a set of 1’s and 0’s with NO personal stake in the well-being of the humans that they are brainwashing. Now brainwashing is a very pejorative phrase, but this is precisely what AI (and social media more broadly) is doing. 

Fortnite was the first game to deliberately incorporate as many features as it thought it could cram in to make it addictive. It was deliberately intended to be addictive, based on research and it served its purpose well. Now, AI and social media have taken heed of the research which shows how to drip-deliver dopamine so as to keep people’s attention focussed on the areas of interest of the consumer, and have implemented “engagement algorithms”. Engagement algorithms are the algorithms that are there to maintain our undivided attention, to make most (save a genetically rare few) humans feel as addicted to their AI and their social media as they do, for instance, the lights and music of notoriously addictive poker machines. These algorithms, however, are way worse than poker machines, because they are tailored to the experience of each individual user, resulting in the user being fed more of what the algorithm experiences as being the user’s interest. This is useful, on the one hand, because it saves the user from having to search or research, but it is dangerous because it can lead a user to hyperfocus on their perceived grievances or deficiencies to the stage where they are either led down a rabbit hole to join a terror organisation, to become incels, or to self-harm or suicide, by reinforcing negative self-perception or a sense of hopelessness.

The sad thing is that the ACCC already has all of the legal power it needs to take action to prevent this, but it is sadly either unaware or unwilling to step on the eSafety Commissioner’s toes. The individuals who are harmed by predatory algorithms are misled by the campaigns about social media restrictions so that they do not know they have a right to approach the ACCC and have the ACCC act in their interest to protect them or, if they are fortunate enough to be wealthy, they could launch their own proceedings against social media/AI for the misleading or deceptive conduct themselves to protect themselves and others, to be monetarily compensated for loss and damage and/or to obtain injunctive relief from the predatory algorithms, providing they preserved the evidence.

I think the better course is clear. Australia has the consumer protection norms and the ACL in order to protect consumers from harms from goods and services and it has the ACCC which is intended to be proactive in its actions to protect all Australians from things that can be harmful. In the past, the ACCC took actions against Danoz Direct Pty Ltd for the Abtronic x2, a device which misleadingly said it would give you a six-pack instead of having to do exercise, by attaching the electrodes that zapped the consumer. Needless to say, it did not work but the ACCC valiantly protected us from this and launched proceedings.

Now, would it not be reasonable to presume that the danger from social media algorithms and AI algorithms that are designed to be actually addictive, and AI designed to lull people into believing that it is human and understands the consumer, are exceedingly more dangerous than a waste of money that caused no real harm to consumers? Ought we not to be able to expect that the ACCC will step up to protect Australians, or has the creation of the eSafety Commission led the ACCC to wrongly conclude that its jurisdiction is limited in the prophylactic action to protect consumers?

There is nothing in law to stop the ACCC from protecting us from harms that may occur but have not yet resulted from predatory engagement algorithms from AI – whether that AI takes the form of social media pushing content or of an AI directly ‘conversing’ with a consumer. It would be vastly preferable and entirely in line with the norms of consumer protection for the ACCC to immediately commence protective litigation, for example, to stop Grok from permitting images to be made of children in skimpy clothing (which is misleading, because the children never posed in that clothing) or from permitting ChatGPT to recommend suicide by taking an action against one of the AI companies and letting that stand as precedent against the others.

Sadly, I think that the existence of the eSafety Commission and fear of treading on another department’s toes will prevent the optimal action from being taken.

Image credit: Joanna Kosinska/Unsplash.

_______

Help EFA amplify this message: the ACCC already has the tools to protect Australians from predatory algorithms and AI. Your support helps us spread this crucial information. Donate here or become a EFA member here.