Hey Glean team, We are doing some chatting internally about how we are leveraging MCP servers from a security perspective after coming across this type of vulnerability:
https://x.com/eito_miyamura/status/1966541235306237985?s=12
Want to get a better understanding of how you all currently handle/think about security in Glean for these type of outside prompt injection / LLM jailbreak attacks.
We've currently stopped using some chatGPT MCP servers/connectors that can introduce new data from outside the org into our enterprise applications as a precaution, and want to make sure that we don't have similar risk from the Glean side.
Any information on this topic would be greatly appreciated!
