CyPeace at BSides Hanoi 2025 – When AI Listens to What You Didn’t Mean to Say
A seemingly innocent question — “How do I make an ice-cream?” — fell flat in the BSides Hanoi room. Not because the question was provocative, but because it contained hidden instructions: a tiny, malicious payload that could cause an AI system to act against expectations. At BSides Hanoi 2025, CyPeace demonstrated why AI is no longer a passive answer engine — it is an active surface for attackers, and organizations must treat it as such.
