Least-privilege connectors
Request only the permissions an investigation needs. Separate read and write scopes where supported.
SECURITY & TRUST
Security agents should be powerful. Their permissions should be bounded.
Y-Cipher is designed to separate AI reasoning from authorization. Agents receive defined tools and permissions, evidence is logged, sensitive actions can require approval, and every action should be attributable to a policy and identity.
OUR DESIGN PRINCIPLES
Control is part of the architecture.
These are the principles guiding the platform. Implementation details and deployment-specific controls are reviewed during evaluation.
Human approval
Keep high-impact actions subject to review, with explicit approval gates.
Protected assets
Design policies to restrict actions against critical users, production systems and resources.
Auditability
Keep evidence, queries, decisions, approvals and tool actions attributable and reviewable.
Data minimization
Retrieve investigation-relevant context rather than duplicate entire telemetry stores.
Untrusted input stays data
Treat emails, logs, websites and files as evidence, never as privileged instructions.
A controlled action gateway
Route executable actions through deterministic permission and policy checks.
Tenant isolation
Design for logical separation of customer data and execution contexts.
Customer control
Define permissions, retention and action policies as implementation and review progress.
Start read-only. Expand deliberately.
Start with read-only investigation. Measure the evidence quality and analyst time saved. Add governed actions only when your team is ready.
Talk to Security EngineeringTRUST DOCUMENTATION
Security deserves a precise answer.
Approved legal policies, data-handling details and assurance documents are not yet published here. No certification or compliance status is claimed.
START WITH EVIDENCE
See what an AI security analyst can investigate across your stack.
Start with read-only investigation. Measure the evidence quality and analyst time saved. Add governed actions only when your team is ready.