Thu 8 Oct 2026 01.00 AEDT

Photo: AAP Image/Sarah Wilson
We’re hearing a great many troubling claims at this week’s public hearings of the parliamentary inquiry into AI, and no doubt there are plenty more to come.
As we follow the proceedings, it’s vital that we stop humanising AI like Big Tech does. Let’s make a start.
When AI developers say that they’re “training” their models, what they actually mean is that they’re feeding them massive amounts of new data. Without an inexhaustible supply of words, images, equations and other cultural and scientific assets, large language models can’t improve their guesses. All they can do is regurgitate patterns based on the data they’ve already got. The global race that AI developers are in is to secure as much of what humanity has already created as possible, and then churn it through their models.
This process, of course, is not training. It’s assimilation – and it’s endless in scope.
At this week’s inquiry, Anthropic claimed that, within current Australian law, it was “technically impossible” to obtain “a licence for the use of every piece of content on the internet as training material”.
Let’s take a step back. Just as every unsecured object in the world isn’t yours for the taking, so too every piece of journalism, photography, literature, research or art on the internet isn’t yours by default. Intellectual property belongs to its copyright holder, no matter where you might encounter it.
Far from being “technically impossible”, Australia already has sophisticated copyright licensing regimes that allow legal uses of music, literature, journalism, research and art – with the consent of copyright holders and with appropriate compensation.
Not wanting to comply with the law isn’t the same as that being impossible.
Just like word and image processing software, AI products are made available to users via licenses. Gone are the days when you downloaded an app or installed it from a CD; Big Tech expects a monthly fee from us, rather than allowing us to own their product – so we know that they understand what licensing is. It’s entirely reasonable, therefore, that Big Tech pays to license the work they hope to assimilate – on the clear understanding that they don’t own that work any more than we own theirs.
AI does not think through a problem or know all the answers. Cory Doctorow’s explainer puts it best: AI is not an intelligence but an unprofitable “word-guessing machine” that generates “an outcome that seems like thought without thinking”.
This “outcome” isn’t journalism, photography, literature, research or art. It’s content: that interchangeable filler that occupies the space and emulates the format of an email, document or image while creating a lot of extra unpaid work for the employees tasked with cleaning it up.
Australian legal giant HWLE recently warned its clients to beware of integrating AI-enabled products into their workflow because “AI behaves in probabilistic and often opaque ways, creating uncertainty and increasing the likelihood of unintended legal exposure.”
Quality research, analysis, explainers and factchecks from experts you can trust.
And yet, all over the world right now, AI developers are jurisdiction-shopping in the hope of securing that one influential nation ready and willing to bend its legislative environment in AI’s favour, then leveraging that precedent across other jurisdictions to reset global frameworks.
What’s at stake if we fold? Weakening Australia’s clear and strong copyright protections “would turn Australia into the data-laundering centre of the world”, warned Copyright Agency CEO Josephine Johnston in response to questioning from Senator David Pocock on Tuesday.
What happens when we give something a human face? We anthropomorphise it. We project characteristics and motivations it can’t possibly have. We distance ourselves from its actions. We diffuse any sense of our own responsibility.
No wonder the White House prefers the term “super intelligence”. As in: “President Trump has taken decisive action to ensure that the United States leads and wins in Super Intelligence”.
When OpenAI apologised only very belatedly for hacking into multiple Australian government systems, they lamented their AI agent’s “misaligned behaviour” and – in the absence of regulatory compulsion – announced a voluntary framework for reporting only the breaches that they choose to publicise.
This came just weeks after the leaders of the world’s biggest AI developers had begun warning that their products risked killing us all.
In the detached language of the casual observer expressing concern, OpenAI CEO Sam Altman argued that “the world should accept some bad things happening for the benefits of this technology and people having the agency.” Rather than accept responsibility for their actions, instead Big Tech has normalised the language of “AI agents” having “gone rogue”.
Even Alan Kohler got in on this action, heralding AI as “the advent of a new life form”.
Let’s be clear. AI is not an intelligence; it’s not being trained, it doesn’t think, it doesn’t have agency, and it’s not alive. It’s a commercial product whose manufacturers insist on humanising in order to bypass laws they’ve already broken and weaken laws they’re keen to flout.
The sooner we stop humanising AI, the sooner we stop dehumanising our present, and the sooner we start rehumanising our future.
AI is just one of many remarkable human creations. Regulating it responsibly will ensure that artists, researchers and journalists can keep creating the work that truly informs and inspires us. In the words of our prime minister: “Anything less, is theft.”
Esther Anatolitis is a writer, creative strategist and national arts leader with more than two decades’ experience shaping Australia’s cultural and public policy landscape.