Sandbox Security Research at Cua
Advancing kernel-level isolation and secure execution environments for AI agents.
At Cua, we're building infrastructure that enables AI agents to interact with computers safely and reliably. As autonomous agents gain greater access to operating systems, applications, and tools, strong isolation boundaries become increasingly important.
Our Sandbox Security Research initiative focuses on strengthening the security foundations of agent execution environments through vulnerability research, offensive security testing, and sandbox hardening.
Security Testing and Responsible Research
Cua Sandbox Security Research is an internal security function dedicated to identifying and mitigating vulnerabilities in Cua's sandbox and agent-execution infrastructure.
Our research is designed around controlled, authorized testing of Cua-owned systems, using isolated environments and documented testing scopes.
Research findings inform security improvements, vulnerability remediation, and regression testing to strengthen sandbox isolation over time.
Research Leadership
Francesco Bonacci
Founder & CEO, Cua AI
Security Research Function Lead
Francesco oversees Cua's Sandbox Security Research initiative, including its research direction, internal testing authorization, and security improvement priorities.
Our Mission
We believe secure isolation is a fundamental building block for trustworthy computer-use agents.
Our goal is to advance kernel-level security, virtualization isolation, and sandbox escape resistance, helping establish stronger security boundaries for the next generation of autonomous AI systems.
Contact: f@trycua.com
Research Focus