Everything Threat Times has reported about AgentCore: 4 stories, newest first. Part of our Cloud Security coverage.
Palo Alto Networks and Zenity Labs report that AWS AgentCore default settings allow credential exfiltration via prompt injection, illustrating broader AI agent risks.
A new attack chain dubbed AgentCorruption allows attackers to hijack AI agents via a single prompt, leading to credential theft and lateral movement within AWS accounts.
Analysis reveals that AWS AgentCore’s security model is undermined by metadata token transmission, weak virtual machine isolation, and overly expansive permissions.
A vulnerability dubbed 'AgentCorruption' in AWS Bedrock AgentCore allowed attackers to compromise entire AI fleets with one prompt.