An Experimental Evaluation of Multimodal Prompt Injection Attacks on Agentic AI Frameworks
This work presents MMPIBench, a reproducible benchmark that measures how far each injected instruction travels through the agent, from perception through planning to the tool call, and extends the benchmark to audio, the only other raw perceptual channel current frontier models accept.
V. Nguyen, Mohammad I. Husain
· 1 citation