AI and data
Lumos integrates artificial intelligence functions for voice transcription, finding suggestions, and writing improvement. This page explains which providers Lumos uses, what is transmitted, and Lumos’s stance on training models with your data.
The AI vendors used
Lumos relies on two third-party vendors for AI functions:
| Vendor | Function | Model | Place of processing |
|---|---|---|---|
| Anthropic Inc. | Text content generation, writing assistance, reformulation suggestions, BNQ Compliance AI Review | Claude | United States |
| OpenAI | Voice recording transcription | Whisper | United States |
Before each data transfer to these vendors, Lumos performed a Privacy Impact Assessment (PIA), per Law 25. You can request a copy of the relevant PIAs by writing to privacy@lumos.property.
The central principle: no training without consent
Lumos does not use your data to train its AI models.
Your reports, voice recordings, photos, and observations are not reused for AI model training, neither Lumos’s internal models nor those of third-party providers like Anthropic or OpenAI.
If Lumos plans to use your data for training purposes in the future, you will be informed with reasonable notice and have the possibility to consent or refuse. Refusing will not affect your access to the Software.
This is a firm stance written into the Privacy Policy (section 4.2) and Terms of Use (section 9.3).
What transits to vendors
To Anthropic (Claude)
When you use AI to generate content (finding suggestions, reformulations, BNQ Compliance AI Review), Lumos sends to Claude:
- The textual content you have entered (description, observations).
- The context of your inspection (section, sub-section, property type).
- For AI Review: the personal findings you have written (Bank findings are filtered out as already compliant).
Anthropic processes the request in real time, returns a response, and does not use your data to train its models under the contractual engagement with Lumos.
To OpenAI (Whisper)
When you use voice dictation, Lumos sends to OpenAI:
- The audio file of your dictation.
OpenAI transcribes, returns the text, and retains the audio for up to 30 days for service abuse detection purposes. Your audio is not used to train OpenAI models.
For audio flow details, see Audio capture consent.
Your responsibility as an inspector
The inspector remains the sole and unique professional responsible for any report produced via Lumos. AI is a writing aid tool. It does not constitute professional, technical, or legal advice.
It is up to you to:
- Verify, validate, and correct entirely the content generated by AI before delivering the report.
- Ensure that the report faithfully reflects the observations made on site.
- Guarantee compliance of the report with applicable standards (BNQ 3009-500 or others).
- Personalize the report based on each building’s specifics.
- Assume the full professional responsibility of the report delivered to the client.
See section 7 of the Terms of Use for details.
Depersonalized real estate data
Beyond real-time AI processing, Lumos also extracts depersonalized technical data from submitted reports (street address, inspected systems, technical observations) to constitute a building condition history by address.
This data:
- Contains no information identifying the inspector, buyer, seller, or any other party.
- Constitutes a platform asset, distinct from Client Data.
- May feed real estate intelligence services by address.
For full details, see section 7 of the Privacy Policy (/legal/privacy) and section 9.4 of the Terms of Use (/legal/terms).
If you want to abstain from AI entirely
Lumos is usable without any AI function. You can:
- Disable voice dictation in Settings > Privacy.
- Not use AI generation buttons in the inspection editor.
- Not enable the BNQ Compliance gauge (which includes an AI Review option).
You retain access to all other features: template structure, narrative bank, photos, PDF generation, delivery, billing.