Professional agents increasingly work through command-line interfaces and multiple tools over shared workspaces containing files, messages, and business applications. These environments create safety risks that are difficult to capture with attacks …
We study multimodal safety evaluation in generative agent social simulations, examining how multi-agent systems built on vision-language models behave under safety-relevant scenarios.
We study how semantic cues can be used to steer safety judgments in vision-language models, proposing SAVeS as a method to improve the reliability of safety-relevant decisions made by multimodal models.