#MLSecOps
"Multi-Agent Systems Execute Arbitrary Malicious Code", 2025.
]-> Hijacking multi-agent systems in your PajaMAS
]-> MAS hijacking tool
// In real-world applications, multi-agent systems will inevitably interact with untrusted inputs, such as malicious Web content, files, email attachments. Using several multi-agent frameworks we show that control-flow hijacking attacks succeed even if the individual agents are not susceptible to direct or indirect prompt injection, and even if they refuse to perform harmful actions
Post #1485
178