Scope enforcement
Target controls and rules of engagement are treated as mandatory execution boundaries, not prompt suggestions.
AI Offensive Security
Human-controlled autonomy for offensive security.
HORUS is an AI-assisted offensive security platform designed for authorized assessments where scope, rules of engagement, sensitive actions and evidence must remain explicit and auditable.
Human approval
Sensitive or high-impact actions can require explicit human-in-the-loop approval before execution.
Auditability
Agent activity, tool execution and evidence are structured to support review, reporting and accountability.
Tool orchestration
Security tooling can be coordinated within bounded workflows for discovery, validation and authorized attack-path exploration.
Provider abstraction
HORUS is designed around multiple AI providers rather than coupling the platform to a single model vendor.
Fail-safe control
Kill-switch, environment policy, sandboxing and explicit approval boundaries reduce reliance on model behavior alone.
Control path
Autonomy is downstream of authorization.
1. Kill switch
A hard operational stop remains available before agent execution proceeds.
2. Scope
Targets are validated against authorized scope before security tooling is allowed to operate.
3. Policy
Environment and tool policy determine which capabilities are executable.
4. HITL
Human approval can gate sensitive actions based on rules of engagement and risk.
5. Evidence
Execution produces traceable technical evidence for review and reporting.
The model can reason about an action. It does not get to redefine the authorization boundary.
Authorized engagements only
Use HORUS to strengthen offensive security workflows without hiding the control plane.
Dewtech can discuss HORUS as a platform capability or as part of a managed, explicitly authorized offensive security engagement.
Start a HORUS conversation