Control is the security feature.

Future ATI is designed around customer-controlled processing, explicit boundaries, role-based access, auditability, and deletion states.

Security Without Theater

Security should be clear and specific. FM does not rely on:

Giant locks
Hacker art
Fake classified stamps
Vague "military-grade" claims

Instead, FM uses system diagrams and clear rules.

Boundary Model

Four zones with explicit rules and access controls:

PUBLIC WEBSITE
No private project files
CUSTOMER ACCOUNT
Billing, scopes, delivery access
PRIVATE WORKSPACE
Files, indexes, prompts, model context
LOCAL DEPLOYMENT
Customer-controlled runtime

Each boundary has enforced rules. Data does not cross without explicit permission.

Access Control Roles

Every private workspace supports authenticated users with role-based permissions:

Viewer

Read results only

Reviewer

Comment and approve

Admin

Manage users and retention

Operator

Run ingestion/evaluation

Supported features across all roles:

  • Authenticated user identity
  • Project membership verification
  • Permission enforcement
  • Session expiration
  • Least privilege defaults
  • Full audit logging

Audit Trail

Every significant action is logged for security review and compliance:

User signed in
File uploaded
Index created
Model loaded
Answer generated
Scope approved
Payment received
Delivery package claimed
Deletion completed

Audit logs do not expose unnecessary private content. They record state changes, not document contents.

Encryption

FM uses encryption where data is stored and transmitted:

In Transit

Encrypted connections where networked. TLS by default.

At Rest

Encrypted storage where private workspace data is retained.

Customer-Controlled Deployments

Storage and key management depend on your selected architecture.

FM does not overclaim before implementation and security review. Your deployment controls the key.

Data Retention

FM explains every retention decision explicitly:

→What is retained (files, indexes, embeddings, logs)
→Why it is retained (active project, delivery, policy)
→Who controls extension (you can extend retention)
→What is deleted (all layers on deletion)
→When deletion begins (immediately after retention ends)
→How deletion is confirmed (backend verification)

Retention is tied to visible state. You always see where your data is.

External Services

FM uses external services only for operational functions that do not touch private data:

Billing
No private project data
Email
No private project data
SMS
No private project data
Account login
No private project data
Public website hosting
No private project data
Analytics
No private project data

Future ATI remains sovereign. External services handle only: billing, communication, authentication, hosting, and analytics. Private documents never pass to these services.

AI Boundaries

Clear rules govern how AI can access and use your data:

External AI
Disabled by default for private work
Local model
Allowed where configured
Tools
Permissioned (must be explicit)
Internet
Policy-controlled

If a tool can write, delete, or access external services, it requires explicit permission.

Security Alerts

When policy is violated, FM alerts with structured information:

Event:BLOCKED
Reason:External AI request denied by policy
Affected:/workspace/session-xyz
Required action:Contact admin to configure external integration
Audit entry:Created

No constant flashing. No vague panic. Alerts explain what happened and what you need to do.

Compliance Language

FM is honest about what it supports:

Allowed

FM can be architected for stronger customer control and network boundaries.

Not Claimed (Unless Proven)

Certified for every regulated environment.

FM does not claim certifications unless they are real and current. Talk to us about your specific requirements.