Understanding capability.
Anticipating risk.
Chrysalia is building security software for AI labs to evaluate AI models in custom cyber ranges. As AI systems become more capable, anticipating security risks requires understanding how they act, adapt and interact with digital environments.
Our research
Our founding team co-authored AI Agents Enable Adaptive Computer Worms, investigating how AI agents can adapt cyberattacks across different targets. The work explores how increasingly capable AI could change the nature and scale of cyber threats.
Read the paper: AI Agents Enable Adaptive Computer Worms (arXiv)
-
“Scientists find way to supercharge dangerous computer ‘worms’ with AI”
-
“…the research is a ‘wake-up call’ to take cybersecurity risks from AI more seriously”
-
“…an AI-driven worm that can’t be stopped by patching a single flaw, because it uses reasoning to detect and exploit different vulnerabilities as it spreads”
-
“This is bigger than Mythos in my view.”
Gary McGraw, Berryville Institute of Machine Learning, in Fortune
-
“This finding catapults us into a new era of cybersecurity.”
Nicolas Papernot, co-author, in The Independent
We’re hiring members of technical staff. To apply or to get in touch, email hello [at] chrysalia.ai.