Topic: langchains response vulnerability

LangSmith Flaw Exposed OpenAI Keys Through Malicious Agent Trick

A critical bug in LangChain’s LangSmith platform allowed malicious agents uploaded to the LangChain Hub to silently exfiltrate OpenAI API keys, prompts, and file data. The exploit, discovered by Horizon3.ai and now patched, underscores growing risks in AI tooling ecosystems where developers freely share unvetted components. Here’s how the attack worked, and what it means for AI security.