The Australian Federal Police have disclosed that children as young as twelve are among dozens of subjects in active counter-terrorism operations since 2020, and that artificial intelligence is dramatically accelerating the radicalisation pipeline. In testimony to a parliamentary inquiry this week, AFP officials described a coming "tsunami" of youth crime driven by the collision of adolescent vulnerability and generative AI's capacity to produce bespoke propaganda, recruitment scripts, and operational guidance at scale.
The warning is not speculative. It is grounded in casework: investigators have observed recruitment cycles that once stretched over months now collapsing into days, as large language models and image generators allow extremist networks to tailor content to individual targets with minimal human effort. A lonely teenager searching for community can now receive personalised ideological grooming within hours, complete with AI-generated memes, manifestos, and encrypted chat invitations.
The compression problem
Traditional counter-radicalisation models assumed a slow arc—curiosity, engagement, immersion, action—that gave families, schools, and authorities time to intervene. AI has compressed that arc. The AFP's testimony suggests that some subjects moved from initial online contact to planning violent acts in under a fortnight. The agency is now lobbying for expanded powers to monitor minors online, a request that will test Australia's already contentious balance between security and civil liberties.
The twelve-year-old threshold is not symbolic. It reflects the practical reality that pre-teens are now fluent in the same digital environments where extremist content circulates. Platform age-verification remains porous, and parental controls are often outpaced by children's technical literacy. The AFP's disclosure is an implicit admission that existing safeguards have failed.
A global pattern
Australia is not an outlier. European and North American agencies have reported similar trends, with the UK's MI5 last year noting a "marked decrease" in the average age of terrorism-related referrals. What distinguishes the Australian case is the explicitness of the AI attribution. The AFP is not merely noting that young people are being radicalised online; it is arguing that generative AI is a force multiplier that demands a distinct policy response.
That response remains undefined. The inquiry has not yet produced legislative recommendations, and the government has been cautious about appearing to criminalise youth or restrict AI tools with legitimate uses. But the testimony lands at a moment when Canberra is already wrestling with broader AI governance—data-centre investment, export controls, content moderation—and may accelerate the push for sector-specific regulation.
Our take
The instinct to protect children from extremism is unimpeachable; the policy challenge is that every proposed remedy—expanded surveillance, platform liability, AI content filtering—carries its own risks to privacy and speech. Australia has historically been willing to accept those trade-offs more readily than its Five Eyes partners. The AFP's "tsunami" language is designed to prepare the public for interventions that, a decade ago, would have seemed unthinkable. Whether those interventions will be proportionate, or merely theatrical, depends on details that have not yet been written.
