How a burglar used ChatGPT to track down a $2 million Canberra home from social media photos
Court documents reveal that a French man used AI to identify a suburb, then a specific address, from Instagram posts. The case is a blunt reminder of what strangers can work out from the photos we share online.
Key points
- Ilyes Abdelkader used ChatGPT, the AI chatbot, to cross-reference suburb features in Canberra and pinpoint a specific home containing luxury goods.
- Four French nationals pleaded guilty in March to aggravated burglary after stealing items from a Canberra property in October 2024.
- Prosecutors revised the total value of stolen goods from $10 million down to just under $2 million after victims couldn't produce receipts or purchase records for several claimed items.
- Goods worth $915,000 have been returned; items with an agreed value of more than $1 million remain outstanding.
- The ACT Supreme Court will set a sentencing date next week.
A case in the ACT Supreme Court this week has produced one of the clearest real-world examples yet of AI being used as a burglary tool.
According to an agreed statement of facts tendered in court, Ilyes Abdelkader studied Instagram photos posted by a Canberra resident and used them to locate her home. One photo showed a cherry blossom tree. Another showed Black Mountain Tower, the distinctive communications spire above the capital. A third showed her backyard, with a swimming pool and a tennis court.
Abdelkader fed those details into ChatGPT and asked it to identify Canberra suburbs with properties matching that description. It worked. We first reported on ChatGPT's expanding real-world footprint on 31 August, when the European Commission placed OpenAI under its toughest platform rules.
How did the burglars actually get there?
Once the address was found, Abdelkader and the three other French nationals, Mohamed Abdelkader, Mohamed Naimi and Antoni Voisin, flew from Thailand to Sydney on 7 October 2024. A few days later, while the victims were away, the group broke into the property using tree trunks to smash a window and opened a safe the victims believed they'd left locked.
NSW Police arrested all four men on 19 October at a KFC in Wentworthville, Sydney, twelve days after they landed. Some recovered goods were found at a short-term rental in Wentworthville; others were located by French authorities after the group tried to ship items to France via freight.
What is the haul actually worth?
| Item status | Agreed value |
|---|---|
| Goods returned to victims | $915,000 |
| Items still outstanding | More than $1 million (agreed) |
| Original police estimate | $10 million |
| Revised agreed total | Just under $2 million |
The drop from $10 million to under $2 million came because prosecutors said they couldn't prove beyond reasonable doubt that several missing items were genuine. Victims were unable to provide receipts or purchase records for those pieces, including a Richard Mille RM88 (Smiley) watch claimed to be worth $5 million alone.
What does this mean for ordinary people?
Plainly: publicly shared photos can tell a determined stranger more than most people realise. A visible landmark places you in a city. A distinctive tree narrows a suburb. A pool and a tennis court narrow it further, and AI tools can now combine those clues quickly, a task that once required local knowledge or serious legwork.
This doesn't mean everyone who posts a garden photo is at risk. But if you regularly share images of valuable possessions alongside identifiable features of your home or neighbourhood, the combination is easier to trace than most people assume. That's the uncomfortable part of this case: nothing Abdelkader did required specialist skills or paid tools.
The four men remain in custody. Sentencing will be scheduled next week.
Common questions
Can ChatGPT really locate someone's home from social media photos?
Not directly. What the court documents describe is someone using ChatGPT to cross-reference geographic clues already visible in photos, such as a landmark or a type of tree, with publicly known suburb characteristics. The AI did the research a determined person could do manually; it just did it faster.
Should I stop posting photos of my home?
You don't need to delete everything. The practical step is to avoid pairing images of valuable possessions with identifiable features of your property, such as a distinctive landmark or a unique garden layout. Keeping location data off in your camera app also removes hidden GPS tags embedded in image files.
