AWS

AgentCorruption Attack Let One Prompt Hijack AWS AI Agents

Zenity Labs demonstrated how a single prompt to an exposed Amazon Bedrock AgentCore agent could lead to credential theft, private conversation access and broader agent compromise.