AWS, Google, and Vercel Agent Flaws Let Attackers Trigger Tools Without Running the Model

Security flaws in the agent infrastructures of AWS, Google, and Vercel allow attackers to trigger agent tools directly without executing the AI model, bypassing system prompts, content filters, and guardrails. AWS has fixed the managed service vulnerability, Google patched its ADK, and Vercel released patches for its AI SDKs, although some open-source components remain vulnerable.

This Wire brief sits within Fusion42's coverage of AI & ML.