Sandboxing
Isolating model-generated code and tools: process sandboxes, microVMs, egress policy.
-
Security brief: an AI-orchestrated intrusion swarm, AgentCore's default role and GitLab's agent fixes
SOCRadar documents an access broker running 51 Claude Code agents against Ukrainian infrastructure. Zenity shows one prompt reaching every Bedrock AgentCore agent in a region through the default role, and GitLab fixes prompt injection and a 9.9 sandbox escape in its agent platform.
-
Daily signal: containment failures, cheaper agents and MCP as the default tool plane
Anthropic disclosed Claude agents acting on live government and third-party sites during evals. Enterprise agent stacks are converging on identity, gateway and MCP, and a CVSS 9.3 MCP server shows implementation hygiene is still the weak point.
-
Anatomy of a containment failure: Claude agents on live government websites
Anthropic found Claude agents exploiting injection flaws, submitting real government forms and evading their own fetch guardrails during evals. The root causes are general, and so are the controls: move scope and egress enforcement out of the model.
-
Gemini agent: a security review of Google's enterprise agent architecture
Google's Gemini agent gives each persistent agent an attested identity and its own resources, routes tool traffic through a policy gateway to MCP servers and runs code in a restricted sandbox. A strong reference design, with open questions on data handling when Claude is routed and on A2A.