Prime Minister Anthony Albanese used his address to the United Nations General Assembly to make an important point about artificial intelligence: the world needs to cooperate and get ahead of the risks of AI.
He pointed to Australia’s world-first social media age restrictions as an example of what can happen when governments are prepared to act rather than wait for others to move first.
His meeting with Apple executive chairman Tim Cook also put children, technology and online safety firmly on the agenda.
The prime minister is right. But if Australia wants to lead the conversation on AI safety, we also need to confront what AI could mean for the radicalisation of young people online.
When I pushed for an inquiry into how young Australians are being radicalised online, gaming platforms, encrypted chats and algorithms were among the starting points.
The Parliamentary Joint Committee on Intelligence and Security is now accepting public submissions into the nature and drivers of radicalisation in Australia, including the role played by technology.
I want to hear from parents, teachers, researchers, tech platforms and young people themselves.
Extremist groups have been using the internet to spread propaganda and find recruits for years. AI has the potential to supercharge that process.
The evidence is already emerging, but it’s important to distinguish what we know from what we’re still investigating.
The clearest example is propaganda. Groups including ISIS and Al-Qaeda have used generative AI tools to produce recruitment material and disinformation. AI can make that content faster and easier to tailor to different audiences.
There’s also growing evidence that AI can give vulnerable people the practical means to act on extremist beliefs.
Tech Against Terrorism, a UN-backed watchdog, tested 27 AI models with more than 2,000 requests based on real terrorist use cases, including instructions relating to explosives and attack planning.
ChatGPT, the most widely used system, refused only 48 per cent of those requests.
Researchers at Cambridge University have separately found Boko Haram operatives used chatbots, including Claude, Gemini, Meta AI and DeepSeek, to help plan attacks and refine bomb-making techniques.
These findings are confronting.
But the harder question is whether a chatbot can radicalise someone directly, the way a human recruiter might, through persistent conversation.
Researchers call this a real and rising risk. AI doesn’t even need to be deliberately misused to cause harm.
Last year, for example, an update to X’s chatbot Grok’s own instructions led it to generate content praising Hitler and using antisemitic tropes, demonstrating how quickly AI systems can produce harmful material.
The director-general of the Australian Intelligence Security Organisation, Mike Burgess, put the broader challenge well in his Lowy Institute address last November: ‘the internet incubates, social media accelerates … artificial intelligence exacerbates.’
Think about the difference between the internet a decade ago and what a young person can access today.
A young person who once needed to find the right forum, the right social media account or the right person willing to talk them into an ideology, may now find a chatbot that never gets tired, never pushes back and can generate content tailored to what they’re feeling in that moment.
Radicalisation is complicated and there are many different pathways into extremism, but that’s exactly why we need to understand where AI fits into those pathways.
Gaming platforms, algorithms and encrypted messaging remain central to how young Australians are being exposed to extremist thinking. AI cuts across all of them.
It can write the propaganda, generate images and videos, produce material shared through gaming-adjacent chats, and have a conversation with a vulnerable young person at a scale no human recruiter could match.
We shouldn’t wait for a young Australian to become radicalised through an AI system before we start asking what safeguards are in place.
We must ask those questions now.
If you work with young people, research this area, build the platforms, or are simply a parent who’s thought about what your children might encounter online, I encourage you to make a submission to the Parliamentary Joint Committee on Intelligence and Security.
The committee wants to hear from the community, not just experts, because this affects families in every part of Australia.
Submissions can be made through the committee’s website. The inquiry will report by 22 March 2027, but its findings can only be as good as the evidence put in front of it.
At last week’s UN General Assembly, the prime minister argued that the world can’t afford to wait for AI risks to emerge before we act. He’s right.
If Australia wants to lead on this issue, that leadership has to start at home, including in the places where young Australians are already learning, playing and forming their views about the world.
Leave a comment