Skip to content

Trust & Verification

How MLNavigator approaches offline review workflows, deployment records, energy reporting, and data handling in high-assurance environments.

Note: This page stays at a public-summary level. Detailed implementation materials are reserved for private review.

Threat Model

What We Protect Against

  • Unauthorized data exfiltration during inference
  • Unexpected changes in approved software or model packages
  • Configuration drift without a review trail
  • Unclear runtime behavior during review or incident response
  • Log deletion or modification

What We Do Not Claim

  • AI outputs are “true” or “correct”
  • Protection against hardware-level attacks
  • Protection against malicious operators with root access
  • Model safety or alignment guarantees
  • Perfect numerical reproducibility across architectures

Field notes from customer and operator interviews:

  • "People want the tech, but they do not trust it yet." — Product stakeholder, regulated enterprise
  • "IT is spearheading AI governance, and teams follow once controls are published." — Technical director, enterprise integrator
  • "Every employee signs our AI directive." — Senior vice president, component manufacturer

Operational Records

The draft spec outlines retained records that support later review of how a run was performed.

The public site keeps this description at a high level: records can preserve deployment context, operational history, and output provenance for review workflows.

Record Scope

Public descriptions focus on categories rather than schemas: deployment context, approved configuration, operational history, and output provenance.

Operating Consistency

Declared operating rules

Public materials focus on the fact that operating rules are documented and reviewable.

Recorded change context

Teams need enough context to interpret changes over time without reconstructing them from scratch.

Review boundaries

Public claims stay at the boundary level: what the records support and where human judgment still applies.

Note: Public descriptions stay at the governance level and avoid implementation-specific thresholds or controls.

Energy Measurement Method

We discuss energy reporting publicly at a high level so teams can compare local deployment approaches without relying on vendor marketing.

Public methodology summary

The public methodology summary stays qualitative: establish a stable baseline, run a representative workload, and report conditions clearly enough for internal comparison.

Required Reporting

  • • Deployment environment
  • • Test window
  • • Operating conditions
  • • Reporting assumptions

Repeatability Rules

  • • Comparable environments
  • • Consistent reporting window
  • • Representative workload selection
  • • Consistent reporting conditions

No Third-Party Collection

Explicit Statement: This website does not include third-party tracking scripts, analytics pixels, or telemetry that reports to external services. adapterOS is designed without them.

We do not use Google Analytics, Facebook Pixel, Mixpanel, Segment, or similar services.

Allowed Minimal Logs

Minimal server access logs for security and debugging. IP addresses, timestamps, request paths. Typically retained for 30 days. Not shared with third parties.

Opt-In Only

Contact form submissions are stored to respond to inquiries. Newsletter signup is explicit opt-in only. No pre-checked boxes.