Xage Security announced new Zero Trust capabilities designed to make AI agents jailbreak-proof through deterministic visibility and control [1].
This development addresses a critical vulnerability in the deployment of autonomous AI agents. As organizations integrate AI into sensitive operations, the risk of "jailbreaking"—where an agent is manipulated to bypass safety protocols—could lead to catastrophic failures in critical infrastructure.
The new solution provides end-to-end visibility and control over AI agents across various environments, including SaaS, cloud, in-house data centers, and the edge [3]. To demonstrate the efficacy of the system, Xage Security showcased the technology by successfully blocking a compromised OpenClaw agent [1].
This launch follows a period of significant expansion for the Palo Alto-based company [3]. Xage Security reported a year-over-year revenue growth of 81% [3]. Additionally, the company saw a 102% year-over-year increase in its number of customers [3].
The company said these capabilities were announced on March 18, 2026 [3]. The tools are intended to enable secure production deployments by ensuring that AI agents operate within strict, predefined boundaries, preventing them from executing unauthorized actions even if the underlying model is compromised [1].
By implementing a Zero Trust architecture specifically for AI, Xage Security aims to move beyond traditional probabilistic security measures. The company's approach focuses on deterministic control, which means the system knows exactly what an agent is permitted to do and can block any deviation in real time [1].
“Xage Security announced new Zero Trust capabilities designed to make AI agents jailbreak-proof”
The shift toward deterministic control represents a move away from relying on the inherent safety filters of Large Language Models, which are often bypassed by sophisticated prompts. By placing a Zero Trust layer between the AI agent and the critical infrastructure it manages, companies can deploy autonomous systems with a reduced risk of systemic failure or malicious takeover.
