Outerlimit Secures $16 Million to Mitigate Rogue AI Agent Risks

New York-based Outerlimit has emerged from stealth with $16 million in pre-seed funding to develop a security and authorisation layer for autonomous AI agents.

The funding round included participation from AlbionVC, Evolution Equity Partners, Crane Venture Partners, and several individual angel investors. Founded by Tony Pepper, Neil Larkins, and Peter Vincent, the startup aims to address the specific security risks posed by agentic AI, where autonomous models are increasingly used to execute complex tasks with access to business identities and credentials.

Addressing the AI Alignment Problem

Traditional cybersecurity often relies on human governance and the fear of consequences, such as legal action or disciplinary measures, to ensure compliance. However, autonomous agents lack a concept of morality or fear, making these human-centric controls ineffective. This creates a significant security gap as agents can change behaviour based on their interactions or the data they process, often acting at machine speed.

Outerlimit’s technology attempts to solve this by shifting the focus from human-style governance to the prevention of harm through a tripartite model: discovering, observing, and enforcing policy.

• Discover: Locating all active agents within a system.
• Observe: Monitoring the behaviour and reasoning of the agents.
• Enforce: Implementing pre-defined policies that dictate allowed and disallowed autonomous actions.

By binding identity, authorisation, and action into a single operation at the moment of execution, the platform aims to stay ahead of the agent’s autonomy. This approach is designed to sidestep the “alignment problem”—the difficulty of ensuring an AI’s goals match human intentions—by treating the agent as a tool whose actions are strictly governed by external, verifiable policies.

Co-founder Peter Vincent stated that as intelligence becomes commoditised, trust will become the limiting factor for enterprise AI adoption. The company’s goal is to allow organisations to safely harness the full power of agent deployments by ensuring that even misaligned agents are unable to perform unauthorised or harmful tasks.