Ultra Security raises $10M to build safety layers for autonomous AI agents

Emerging from stealth

Ultra Security, a cybersecurity company based in Birmingham, Michigan, has officially launched after a year of stealth development. The company secured $10 million in seed funding to build a security platform specifically for autonomous artificial intelligence systems. This funding came from a group of investors including Ludlow Ventures, Hidden Capital, Entrée Capital, Detroit Venture Partners, firedrop, and FDVC.

The startup was founded in 2025 by Chase Lee, who previously served as the founder and CEO of Trustpage. Trustpage was later acquired by Vanta, a company that specializes in security and compliance. Lee's background in enterprise security technology is central to the company's goal of securing complex AI infrastructures.

The company is positioning itself to meet a growing demand for security tools that can handle the unique risks of AI agents. Unlike standard software, these agents can independently access applications, process sensitive information, and execute various tasks. Ultra Security aims to provide the governance needed as organizations begin to deploy these more capable systems.

How the platform works

Ultra Security's platform operates at the intersection of AI agents and the Model Context Protocol (MCP) servers and connectors. It works by tracking every request for data, every prompt, and every tool call as they happen during an active session. This allows companies to oversee and manage what an agent is doing as it happens, rather than just looking at logs after the fact.

The technology focuses on four primary areas: guardrails, governance, observability, and connectivity. The guardrails allow companies to turn plain-language policies into real-time actions, such as blocking access to credentials or redacting personally identifiable information. It can also prevent unauthorized data transfers between different connected systems or limit abusive activity.

For administrators, the governance features determine which specific agents, tools, and connectors are available to different employees or teams. The observability component logs and traces every interaction between an agent and a tool. It also uses AI-native anomaly detection to identify suspicious behavior and generate alerts for the security team.

Architecture and compliance

A key differentiator for Ultra Security is its local-first architecture. This design ensures that tool calls, payloads, and audit records remain on an organization’s own devices by default. While this provides a high level of privacy, the company also offers a remotely hosted deployment option for organizations that prefer it.

The company is also emphasizing its commitment to industry standards and compliance. Ultra has secured certifications for SOC 2 Type II, ISO 27001, and ISO 42001. They have also reached AARM Extended conformance and aligned their security framework with the NIST AI Risk Management Framework and Agentic Trust Controls.

To support government and highly regulated sectors, the company has finished a FedRAMP 20x Class B audit conducted by an outside assessment firm. They are currently in the process of pursuing full FedRAMP certification. These credentials suggest a focus on enterprise-grade security and high-stakes environments.

Closing the security gap

Traditional cybersecurity tools like web application firewalls, API gateways, and cloud security posture platforms were not built for autonomous AI. These legacy systems often struggle to monitor the complex, multi-step actions that occur during an active AI agent session. Ultra Security is attempting to fill this gap by combining runtime protection and policy enforcement into one environment.

The platform is designed to be flexible enough for organizations with varying levels of AI maturity and technical expertise. Whether a company is just starting to explore AI or has a complex infrastructure, the tool aims to provide a unified security layer. It handles the complexity of agents interacting with email, software development environments, and other business-critical systems.

By offering a universal connectivity layer, Ultra Security aims to work across different AI agents, models, and connectors. This means organizations do not have to create separate configurations for every different client they use. The goal is to provide a scalable way to manage the security of an expanding AI ecosystem as it becomes more integrated into daily business operations.