NEW York-based Outerlimit has emerged from stealth with $16 million in pre-seed funding from AlbionVC, Evolution Equity Partners, Crane Venture Partners and angel investors. Founded by Tony Pepper, Neil Larkins and Peter Vincent, the company is developing a decentralised security and authorisation layer for autonomous AI agents.
The company argues that enterprises are giving agents increasingly broad access to business identities and credentials while using them for more complex decisions and actions. Because AI models operate probabilistically and agents can change their behaviour in response to information or other agents, it is difficult to ensure that their actions remain aligned with their intended purpose.
Outerlimit says conventional controls based on human rules and consequences are ineffective for autonomous software, which has no understanding of morality or fear of punishment.
Its approach is built around “discover, observe and enforce”: locating agents, monitoring their behaviour, and applying policies governing permitted actions. Outerlimit says its technology can prove that delegated tokens will be used only within specified scopes, locations and conditions. This is intended to make the policy, rather than the agent’s alignment or developer instructions, the controlling factor.
The firm claims it can operate at machine speed and block harmful actions before they occur, although the article provides no independent evidence or confirmed incidents demonstrating those capabilities.