
Vulnerabilities
AWS patches Bedrock AgentCore SDK flaw allowing sandbox command injection
A remote authenticated user could run arbitrary commands in the Code Interpreter sandbox via crafted package names in older SDK versions.
Everything Threat Times has reported about Bedrock AgentCore: 2 stories, newest first. Part of our Vulnerabilities coverage.

A remote authenticated user could run arbitrary commands in the Code Interpreter sandbox via crafted package names in older SDK versions.

A vulnerability in the Bedrock AgentCore Python SDK exposed sensitive AI prompts and responses in CloudWatch logs until version 1.5.1 was released.