Ultra Security is an artificial intelligence (AI) security startup based in Detroit, Michigan. Founded by a team of data, network and security professionals, Ultra Security aims to enable fearless innovation through artificial intelligence. Its product, Ultra, is an enterprise-level security platform that governs AI agents.

Companies worldwide are beginning to or already using AI agents daily, and these agents often access sensitive data. However, many companies either don’t know the risks of providing agents with this access or don’t know how to begin leveraging a secure AI agent in their workflow.

The Ultra software suite mitigates these risks, though the process and benefits can be difficult to grasp for prospective users who may be unfamiliar with agentic workflow semantics.

Our Sandbox for Neutralizing Agentic Risks & Exploits (SNARE) solves this issue by providing a safe, isolated environment for an agent to perform potentially unsafe actions.

Users are presented with storylines that display unsafe agent behavior that leads to potential security risks. By comparing the outcome with and without Ultra’s guardrails, prospective users are better informed of the agent’s capabilities.

SNARE interacts directly with the Ultra dashboard to show how specific unsafe actions by the agent are neutralized by Ultra’s guardrails. Using this integration, clients are shown these interactions that normally happen out of sight. The storylines SNARE generates enhance the user’s understanding of agent interactions through instant feedback.

SNARE runs in Docker containers, where Ultra software governs a self-hosted agent. The agents use a custom Model Context Protocol server. A Flask back end interacts with a Node.js environment containing a React and Next.js-based front end.