
Upwind launched an AI inventory layer that maps relationships, dependencies, and risks across cloud AI services. It monitors prompts, tool calls, MCP interactions, and data retrievals as exposure points. The system detects prompt injection, over-permissioned tool calls, and data leakage in real time. It supports AWS Bedrock, Azure AI Foundry, Vertex AI, self-hosted models, and custom agents.
Tap to vote and see what everyone thinks.
Models Not Primary Source of Cybersecurity Risk
Summary by ByteBrief