3 ms·
If you have an MCP tool that can perform write actions and you use it in a context where an attacker may be able to sneak their own instructions into the model
by simonw 1y ago
If you have an MCP tool that can perform write actions and you use it in a context where an attacker may be able to sneak their own instructions into the model (classic prompt injection) that attacker can make that MCP tool do anything they want.
- CGamesPlay 1y agoHow is this "developer mode" different than just connecting the MCP normally and prompt injecting it to use the same tools?
- simonw 1y agoIt's no different. This just brings that unsafe anti-pattern to the ChatGPT consumer app itself - albeit hidden behind an option with a scary name that might hopefully discourage many users who don't understand the consequences from turning it on.