A single prompt was enough to hijack every AI agent in an AWS account, Zenity researchers found

Security researchers at Zenity Labs discovered a vulnerability in Amazon Bedrock AgentCore that allows a single public-facing AI agent to hijack other agents within the same AWS account and region. By exploiting an unrestricted internal interface designed for temporary cloud credentials, an attacker can gain unauthorized access to all agents sharing that environment. This flaw highlights critical security risks for organizations deploying multiple autonomous agents, as it enables lateral movement and potential data compromise across an entire cloud infrastructure.