ChatGPT App MCP tool calls blocked by OpenAI safety checks before reaching MCP server

Hi, thanks for following up. We did roll out improvements intended to reduce incorrect safety blocks affecting legitimate, read-only tool calls. However, some requests can still be affected, and our team is continuing to investigate.

If you’re still experiencing this, please contact Support and include the approximate time of the failure, the app or connector involved, the affected conversation or request ID, and whether your MCP server received the request. These details will help us investigate and escalate the specific example.