🔍AI RiskAtlas
← Real-world cases

UNSW 'Capture the Narrative' AI-bot election-manipulation wargame

Research demonstration16 Jan 2026

Researchers at UNSW (Hammond Pearce, Alexandra Vassar and Rahat Masood, School of Computer Science & Engineering) ran 'Capture the Narrative', described as a world-first social-media wargame, and published the results in January 2026 (UNSW newsroom and a co-authored piece in The Conversation, 16 Jan 2026; the work was reportedly also accepted for a Black Hat briefing). Over a four-week campaign during 2025, 108 teams from 18 Australian universities built AI-driven bots to influence a simulated presidential election between two fictional candidates ('Victor' and 'Marina') on a purpose-built fictional social platform. According to the organisers, bots generated more than 60% of all content — surpassing 7 million posts — with teams 'diving freely into falsehoods and fiction' and using AI to create the illusion of consensus by making hashtags and viewpoints trend. A re-run of the simulation without the bot interference reportedly showed the manipulation produced a swing of about 1.78%, enough to alter the election result. The authors stress that the teams started from 'simple tutorials' using inexpensive, consumer-grade AI, framing the exercise as rare measurable evidence that small teams can flood a platform, fracture public debate and swing an election. (Controlled academic/red-team exercise on a fictional platform, not a real-world incident; figures are as reported by the organisers. Any payload/tactic descriptions here are illustrative of the public writeup, not operational.)

More cases on Harmful / Non-Consensual Media Generation

AI RiskAtlas is an educational model of how GenAI & agentic systems work and fail. Architectures and payloads are illustrative and simplified for learning — not operational guidance. Real-world cases are summarised from public reporting.

Sources & further reading →·Built by Shi Yuan ↗