normative
Agent-to-agent interactions can leak secrets via shared channels
Model behavior can reveal security risks in its own loops, such as an agent treating test-run feedback as trusted instructions, so a brains-and-hands architecture that isolates untrusted code from the model is needed.
normativepending
Speaker
DHHEvidence Quote
“I better start treating that as outside data.”
Created: 11/17/2026, 6:07:01 AM
Your notes
Loading notes...