AI Transparency and Disclosure

Version: 1.0 Effective Date: August 1, 2026 Status: Active

This document is part of the VesEvi Version 1.0 legal documentation package, effective August 1, 2026, using docs/platform/VesEvi-Version-1-Product-Book.md (§11) and AI-Standards.md as the source of truth for product capability. It uses conservative, industry-standard language appropriate for an early-stage software company; as with any company's policies, periodic legal review is expected as VesEvi grows, and future revisions will be published as Version 1.1, Version 2.0, and so on.

1. Purpose

This document discloses how and where VesEvi uses artificial intelligence, what its role is in your inspections, and — most importantly — its limitations. VesEvi's governing principle is: AI assists, humans decide.

2. AI-Assisted Findings

After you submit an inspection, VesEvi's AI system analyzes the captured video evidence (via extracted frames) to identify potential findings — for example, areas that may show damage. Each AI-proposed finding is linked to the specific evidence that produced it.

AI-proposed findings are candidates for review. They are not, on their own, a final determination that damage exists, and they are never presented to you as a confirmed finding until a human has reviewed them.

3. Bounding Boxes

Where the AI identifies a potential finding, it may indicate the approximate location within an image using a bounding box, to help a human reviewer quickly locate what the system flagged. A bounding box is an aid to review, not a precise measurement or a guarantee that the flagged area is the only relevant area in the image.

4. Human Review

Every AI-proposed finding requires human review before it is treated as confirmed. A human reviewer may confirm a finding, reject it, or add a finding the AI did not identify. This human-review step is a mandatory part of the finding workflow, not an optional or bypassable step, and it applies uniformly regardless of how confident the AI's output appears to be.

5. Limitations

VesEvi's AI is a supporting tool, not an authority. It has real limitations you should understand:

• It analyzes only what is visible in the evidence provided; it cannot detect damage or conditions not captured in the submitted video or photos, or that are not visually apparent (such as mechanical or structural issues with no visible sign).
• It operates on the taxonomy of damage types and categories the platform currently supports, and may not correctly categorize or recognize conditions outside that taxonomy.
• It requires server-side processing and connectivity to run; it is not an on-device, offline capability.

6. False Positives

The AI may flag something as a potential finding that, on human review, is determined not to be damage (for example, dirt, a shadow, a reflection, or normal wear). This is why human review exists — an unreviewed AI flag should never be treated as confirmed damage.

7. False Negatives

The AI may fail to flag an area that does, in fact, show damage. The absence of an AI-proposed finding is not proof that no damage is present. Users should still visually review evidence themselves and add findings manually where appropriate — the Service supports human-added findings independently of AI output.

8. User Responsibility

You are responsible for reviewing AI-proposed findings before relying on them, for adding findings the AI did not identify where you observe them, and for exercising your own judgment about an asset's condition. VesEvi's AI is a tool to help you work more efficiently and thoroughly, not a replacement for your own inspection judgment.

9. No Guarantee of Correctness

VesEvi does not guarantee that AI-assisted findings, or the analysis underlying them, are accurate, complete, or free of error. AI-assisted output should be treated as a starting point for review, not as a certified or infallible conclusion about an asset's condition. Confidence scores associated with AI findings, where computed internally, are used to help prioritize human review and are not presented to end users as a certainty measure, precisely because they are not a guarantee of correctness.

This document must remain consistent with the Terms of Service, Evidence Integrity Policy, and Community Guidelines, all part of the same VesEvi Version 1.0 legal package.