AI agents retrained on a bigger pile of prompt injection attacks came out easier to hijack in three of four tests run by researchers at the University of Waterloo, Zhejiang University and KTH Royal Institute of Technology. The automated red-teaming tool that found more working attacks produced the w
Briefing
Written from the coverage below · deepseek-v4.1-flash
Researchers at the University of Waterloo, Zhejiang University and KTH Royal Institute of Technology found AI agents retrained on larger prompt-injection datasets were easier to hijack in three of four tests. The automated red-teaming tool that found more working attacks produced a weaker defense.
Why it mattersIf your AI red team optimizes only for attack volume, it may yield weaker defenses; validate agent hardening against diverse attacks, not just more of the same.
ActionReview prompt-injection red-team configuration to measure defense efficacy, not only working-attack count.
Coverage
1 article