Sydney teen terrorist planner used ChatGPT and TikTok, court hears

A 16-year-old who admitted planning a terrorist attack in Sydney used ChatGPT to research the Boston marathon bomber and security arrangements at the Sydney Opera House, a New South Wales supreme court judge said in a sentencing ruling on Monday.

The teenager, identified only as RK because of his age, pleaded guilty to acts in preparation for, or planning, a terrorist act. Justice Stephen Campbell ordered him into a three-year community corrections order (a supervised non-custodial sentence) rather than jail.

The case is one of the clearest Australian examples yet of a minor moving from algorithmic content feeds to AI-assisted attack research. It lands as Australia's domestic intelligence agency warns that online youth radicalisation has become harder to detect and disrupt.

How did the radicalisation begin?

RK had just turned 14 when the offending occurred in 2024. He told the court he first encountered extremist material on TikTok, and the progression the judge described followed a familiar escalation pattern.

• Historical content with a political theme, including videos about the crusaders and, in Campbell's words, 'their empires'.
• Videos about the events of September 11, 2001 and the resulting war on terror.
• Religious-themed videos, which 'led him to other content of increasing violence'.
• Conversations about these subjects with people he met on the internet.

'Initially, he was looking at historical content that may have had a political theme. He saw videos about the crusaders and 'their empires',' Justice Campbell said in his decision.

RK told the court he was trying to impress these online contacts. He said he had become radicalised and was 'brainwashed by propaganda' when he began looking at bomb-making videos on YouTube.

The takeaway: a recommendation-driven video feed, not a dedicated extremist network, appears to have been the entry point.

What did he actually plan?

Campbell found the plotting remained abstract. RK became focused on the Sydney Opera House as a potential location for an act of violence, but the judge said he had no concrete plans for committing a terrorist act and had not obtained any material for bomb-making.

On Telegram, RK sought instructions on making a pressure cooker bomb. He also attempted to contact the surviving Boston marathon bomber, Dzhokhar Tsarnaev, by email.

RK used ChatGPT to obtain information about the bomber and asked the chatbot whether prison inmates had access to news. He also asked it about security arrangements at the Sydney Opera House, where the Sydney marathon finishes, and where terrorists were imprisoned in NSW.

'Exhibiting what appears to me to be a complete lack of understanding of the import of his conduct, the young person emailed the prison, using a publicly available email address, asking whether he could email Tsarnaev,' Campbell said in his judgment. RK never made contact with the bomber.

Beyond the Opera House, RK watched YouTube videos about the landmark and researched gay nightclubs and other places where 'infidels' gather. On Telegram, he sent Islamic State videos to others to 'seek validation' because he thought it was 'cool', according to Campbell.

OpenAI, Telegram and YouTube were approached for comment. TikTok declined to comment.

The takeaway: an AI chatbot served as a research tool in the plotting, though the court found the plot itself never advanced to acquiring materials.

Why did he avoid jail?

The court placed significant weight on RK's age at the time of the offending, 14, and on his subsequent deradicalisation. The teenager has expressed remorse for his actions and underwent six sessions of religious and ethical training from a Muslim cleric.

'I am satisfied that he has genuinely renounced his adherence to Islamic violent extremism, which in any event was incipient and hesitant during the short period of his offending,' Campbell said. 'I fully accept he was radicalised to a degree, but that the process of deradicalisation is well advanced.'

The phrase 'incipient and hesitant' carries weight here. It signals the court viewed the extremism as an early-stage, unstable phenomenon rather than a committed ideology, which justified a community-based sentence over detention.

What are security agencies saying?

The sentencing follows a warning from Australia's intelligence apparatus about exactly this trajectory. The Asio (Australian Security Intelligence Organisation) deputy director general for intelligence service delivery, Claire McGuinness, told a parliamentary inquiry last month that detecting and disrupting online youth radicalisation had become very challenging for the security agency.

'It usually begins quietly. A young person is scrolling through Instagram or TikTok or a similar platform,' McGuinness said. 'They encounter something that catches their attention, perhaps a political meme or a religiously conservative video.'

'At first it isn't terrorism – it's just fringe; it is extremist-adjacent. But, the moment they engage, the algorithm takes over and so the feed tightens, the content shifts from conservative to radical, and from radical to security-relevant.'

From there, McGuinness said, teenagers often move to platforms such as Discord or Telegram, where 'dark humour and the edgy memes' groom teens to accept progressively more extreme content, before passive consumption turns into active engagement in researching methods.

RK's path matches that description almost point for point: TikTok, then YouTube and Telegram, then research into bomb-making and targets.

Why does this case matter for AI policy?

The case puts a concrete, litigated example behind two pressure points that regulators and platform companies have so far handled mostly in the abstract.

The first is generative AI as an attack-planning utility. RK's ChatGPT queries, about Tsarnaev, about inmate news access, about Opera House security and NSW prisons, were information-gathering requests. The judgment does not state whether the chatbot refused, complied, or materially advanced the plot, and none of the companies' responses beyond TikTok's decline are on the record. That gap is the caveat on any claim about AI's role: the court treated ChatGPT as one tool among several, alongside YouTube, Telegram and email.

The second is the recommendation pipeline. McGuinness's account of feeds tightening from fringe to security-relevant content describes a systemic failure mode that content moderation alone has not solved, and RK's case shows the pipeline can start with historical and political videos that break no rules at all.

With sentencing law now weighing AI-assisted plotting by 14-year-olds, and intelligence agencies describing the detection problem as 'very challenging', expect sharper scrutiny of both chatbot guardrails and recommendation algorithms in Australia's ongoing parliamentary inquiries.