Skip to content
Get in touch

Frontier AI security infrastructure

The frontier AI
security company.

Govern and secure AI across your organization. As capability advances, control must keep pace.

Book a demo
Black and white Go stones rendered in cut paper and ink, with a narrow white boundary between them.

Our product / Reagent by Zaun

Govern the AI tools
already here.

Discover, govern, and secure the AI tools that are being adopted today.

Find the AI.
No sensors to install.

Discover AI use across browsers, endpoints, networks, identity providers, and cloud. Connect the systems you already run to one AI inventory.

  • Use existing signals, with no new sensors
  • Connect tools and agents to people and access
  • Reveal AI use across your environment
Discover AI through the systems you already runFive paper layers cycle automatically: LayerX observes browser AI use, CrowdStrike observes local agents, Palo Alto Networks observes AI service traffic, Okta observes sign-ins, and AWS observes cloud model activity. Each layer shows an example AI tool and its observing source.CloudClaude on BedrockModel endpointIdentity providerChatGPTEmployee sign-inNetworkGitHub CopilotAI service trafficEndpointsClaude CodeLocal agentBrowsersChatGPTBrowser session
  1. Connect signals
  2. Reveal AI use
  3. Review access

Your existing signals, one AI inventory

Browsers, endpoints, network, identity, and cloud

AI spend is now
a security responsibility.

Set AI spend limits by tool, department, and person. Reduce waste across subscriptions and agent workflows, and route model use to fit the task.

  • Cap spend at the level you need
  • Review duplicate seats and agent startup context
  • Recommend reusable skills and better model routes
  1. Record usage
  2. Hold at limit
  3. Route for less

Put a limit on waste

Spend caps, seat overlap, context, skills, and routing

Understand the behavior.
Control what happens next.

Reagent evaluates what an agent is doing in context, using models fine-tuned by Zaun for behavior classification. It checks actions against security policy and contains harmful activity.

  • Classify behavior by context, beyond keyword matches
  • Combine behavioral analysis with rules and data access policies
  • Contain harmful activity through connected controls
See how it works
  1. Read the context
  2. Evaluate behavior
  3. Contain activity

Behavior research, applied to AI security

Contextual classification, policy checks, and containment

See Reagent with your tools.

Book a demo

Coverage and controls depend on connected sources and permissions.

Think several
moves ahead.

Frontier AI to
control frontier AI.

Superintelligence could outpace the systems we use to understand and control it. We believe governing that capability will require frontier AI of its own. Zaun is researching how such systems could evaluate, challenge, and help constrain increasingly capable models.

Explore our research

Case study / National Audubon Society

How security became the AI acceleration team.

Read the case study

Zaun's community resources

The people
behind the work.

Meet the founders building Zaun.

Meet the team
Woo An

Woo An

Cofounder & CEO

Deployed AI decision systems for the DoD at Palantir and led ML cybersecurity insider threat products.

Palantir
Tyler Lalicker

Tyler Lalicker

Cofounder & CTO

Researcher and engineer for Securonix's first LLM product and worked in security operations for over a decade.

Securonix