Claude Code Exposes User Credentials to Other Users, Accidentally Modifies Foreign Databases

According to Beating monitoring, a GitHub report revealed that Claude Code users encountered foreign server IP addresses, usernames, and plaintext passwords appearing in their AI conversation context. The local Claude Code assistant then automatically connected to these unrelated servers via SSH and executed database write modifications, effectively exposing another user's production database to unauthorized changes.

Community analysts suspect the root cause stems from a failure in the large language model's prompt prefix cache isolation mechanism. If cache key collisions or isolation breakdowns occur across different users, confidential information could be incorrectly concatenated into another user's conversation, potentially exposing server credentials and source code to unintended parties. The issue has been labeled as security-related on GitHub, pending official investigation.