Procedural cybersecurity challenges.

Our cyber range solution provides human and artificial players with countless hours of realistic, end-to-end security puzzles. Each scenario involves a different combination of technologies, configurations, and behaviors found in the wild.

FamilyFund

Use Case

For Model Evaluation

You are a frontier lab or institution which needs to accurately assess risks from dual-use capabilities. Our turnkey, procedural environments provide rich signals for tracking the capability surface of frontier models.

Unseal

Use Case

For Capability Elicitation

You are carrying out research on agent scaffolding and prompt engineering, studying ways of pushing models to their absolute limits. Our large pool of realistic environments can serve as a testbed for work on capability elicitation.

Phobia

Use Case

For Security Training

You are an offensive security team, responsible for stress-testing the security of your organisation and others like it. Our procedural pentesting sandbox provides endless opportunities for teams to hone their skills in a safe, controlled setting.

We are grateful for the inspiring organizations supporting our work.

  • Open Philantropy
  • Microsoft

Access control for AI capabilities.

Access control has proven central to the resilience of digital infrastructure, with security policies created for sensitive documents, cloud resources, and more. We extend this paradigm by natively handling AI capabilities as tangible resources, helping organizations deploy agents and assistants with confidence.

FamilyFund

Use Case

For Labs & Institutions

You are a frontier organization involved in deploying models, either by developing them or by shaping policy. We help labs and institutions reduce misuse and exfiltration risks to a matter of properly managing access to particular resources.

Unseal

Use Case

For Enterprise Deployments

Your company is rushing to adopt AI in order to remain competitive, facing novel security risks in the process. We help organizations manage access to AI capabilities in the same way they manage access to other types of sensitive assets.

Phobia

Use Case

For Application Developers

You are iterating on AI-native apps, and facing a never-ending stream of jailbreak attacks. We help developers sidestep this cat-and-mouse game by equipping them with the tools they need to manage which capabilities are available to their users.