Multi-agent AI systems are more vulnerable to attacks than single agents in most tested designs, with attack success rates varying up to 3.8 times depending on how roles, communication, and memory are structured.
Local” exposes the agent’s own past reasoning; “Shared
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.MA 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
Architecture Matters for Multi-Agent Security
Multi-agent AI systems are more vulnerable to attacks than single agents in most tested designs, with attack success rates varying up to 3.8 times depending on how roles, communication, and memory are structured.