Critical 'BadHost' Vulnerability Puts Millions of AI Agents at Risk

This discovery highlights a significant structural risk in the AI ecosystem, where widespread reliance on core Python libraries like Starlette creates a massive, singular attack surface. With sensitive credentials at stake in MCP servers, this 'trivial to exploit' flaw necessitates an immediate shift toward more rigorous dependency management in AI development.