Zivis jobs and careers

About Zivis

Zivis is a red teaming firm operating on a one-human, multi-agent model. We don’t just assess AI risk—we actively try to break your systems. Our work is hands-on, malicious, and adversarial. We simulate real-world attacks against LLMs, autonomous agents, and AI-powered apps to uncover how they can be manipulated, misused, or exploited. Jailbreaks. Leaks. Confused agents. Prompt abuse. We find the cracks—then show you how to close them. Whether you're shipping a product, deploying internal tools, or just starting to explore AI, Zivis gives you a direct line to how it actually behaves under pressure. What we do: Hands-on AI red teaming Prompt injection and jailbreak attacks Agent misalignment and manipulation testing Data leakage and privacy breach simulation Custom adversarial scenario design The name “Zivis” comes from the Latin civis, meaning “citizen.” Because every AI system eventually meets the public—unexpectedly, imperfectly, maliciously.