ZeroClaw - Zero overhead. Zero compromise. 100% Rust.(github.com)
github.com
ZeroClaw - Zero overhead. Zero compromise. 100% Rust.
https://github.com/zeroclaw-labs/zeroclaw
2 comments
I looked at it briefly and all their talk about security is about things like gatekeeping who gets to talk to the bot over Telegram. They've done literally nothing about the actual problems such as the lethal trifecta. The fact that they don't even acknowledge that made me tag the project as naive.
Does anyone with more expertise than me have an opinion on this one? I'm especially worried about exfiltration or other negative behaviours, where, assuming the server itself is secure from traffic from remote attacks (quite an assumption) AIs themselves may be externally prompted or led towards poor behaviour that hurts the person running ZeroClaw.