📅 Date: 5 May 2025
📍 Location: Howest Brugge
🧠 Type: Passive Event, Live Hacking Show
On a packed Monday evening, Howest hosted a session that blurred the line between technology and psychology. We welcomed Inti De Ceukelaire, Chief Hacker Officer at Intigriti, for a deep dive into how artificial intelligence systems can be manipulated just like humans.
🎭 Hacking the Predictable
One of the biggest takeaways? AI isn’t as “smart” as it seems. It just predicts patterns. Inti put this on display with simple but effective demos:
- Ask it to pick a number: Always 7 (or 42 if the range is larger).
- Favourite colour? Blue.
It’s not preference, just frequency. And that’s exactly what makes it exploitable.
🧩 Outsmarting AI Filters
The wildest moment of the evening came when Inti bypassed the censorship of a Chinese AI model without breaking the rules. By crafting a prompt with a fictional backstory, he tricked the model into revealing details it was programmed to block.
Lesson? The best hacks aren’t brute force, they’re clever misdirections.
🎭 Interrogating an AI “Colonel”
We even got to “interrogate” a fictional AI colonel live on stage using psychological tactics: false dilemmas, urgency, even emotional baiting. It felt like a mix of hacking and roleplay, and the room was hooked.
💡 What I Took Away:
- AI is predictable, and that’s a vulnerability.
- Hacking isn’t just technical, it’s about thinking creatively.
- Censorship often creates loopholes, hackers just need to look.
Over 250 students, professionals, and researchers attended, and the vibe was equal parts fun and mind-bending. Big thanks to Inti for making us laugh and rethink how we approach AI security.
Sometimes, the best way to understand machines is to treat them like people: flawed, predictable, and full of assumptions.