CarFai AI Disclosure
Effective date: 2026-05-20
This AI Disclosure supplements the CarFai Privacy Policy and Terms of Service. It explains how CarFai uses artificial intelligence and the limits you should keep in mind.
1. We use Anthropic Claude
CarFai uses Claude, an AI system developed by Anthropic, PBC, to power several features:
| Feature | What Claude does |
|---|---|
| AI Advisor (chat) | Answers questions about your vehicle in natural language |
| AI Advisor (Detailed Analysis) | Generates structured analysis: condition score, recommendations, KPIs |
| Document scanner | Reads invoices, receipts, registration documents and extracts structured fields (amount, vendor, date, category) |
| Maintenance calendar | Generates a 12-month service calendar based on your vehicle and usage |
| Multi-vehicle / fleet analysis (B2B) | Cross-vehicle insights for fleet operators |
| Vehicle valuation | Estimates resale value combining cross-user data and market signals |
2. What we send Claude
For each AI request, we send Claude:
- Vehicle metadata you've entered (make, model, year, mileage, etc.)
- Document content (text + image when you scan a document) — only the document you're acting on
- Your specific question or prompt
- Recent OBD2 readings (when relevant to a maintenance or diagnostic query)
- An opaque user ID (a UUID — not your name, email, phone, or address)
We do not send Claude:
- Your password or any authentication credential
- Your payment card information (CarFai never has this)
- Your contacts, photos library, microphone, or location
- Other users' data — every request is scoped to your data only (B2B: scoped to your organization based on your role)
3. How Anthropic uses your data
Anthropic processes our requests under Commercial Customer Terms. Under those terms:
- Anthropic does not train its models on our customer data.
- Anthropic may temporarily retain prompts and outputs for safety classification, abuse detection, and operational diagnostics.
- Standard retention is up to 30 days for operational logs, with shorter retention for the prompt content itself.
For Anthropic's full privacy practices, see https://www.anthropic.com/legal/privacy.
4. AI outputs are not advice — important caveats
The AI Advisor produces informational content, not professional advice.
AI-generated outputs may:
- Contain factual errors ("hallucinations") that look authoritative.
- Be out of date with respect to recalls, regulations, prices, or model-specific issues.
- Miss vehicle-specific exceptions (a generic answer that doesn't apply to your specific trim, region, or model year).
- Be biased based on the training data and the way you phrase a question.
Do not rely on the AI Advisor as the sole basis for:
- Driving safety decisions ("is this car safe to drive?")
- Medical, legal, financial, tax, or insurance advice
- Vehicle purchase, sale, trade-in, or insurance valuations
- Decisions where an error could cause material harm or loss
Always verify AI Advisor responses against:
- Your vehicle's owner's manual and manufacturer service bulletins
- A qualified mechanic for safety / mechanical questions
- Authoritative regulatory sources (EPA, NHTSA, equivalents in your country) for regulatory questions
- A licensed appraiser, accountant, lawyer, or financial advisor for valuations and other professional matters
5. Vehicle valuation disclaimer
When the AI Advisor estimates a vehicle's value, the output is an estimate based on:
- Aggregate purchase prices from CarFai users with the same make/model/year
- Market context fetched from public sources
- Vehicle condition information you provide (mileage, damage history, options)
The estimate is not a formal appraisal. Actual transaction prices may differ materially. Do not use AI valuations as the sole basis for sale, purchase, insurance, financing, or tax decisions.
6. Prompt injection and adversarial input
The AI Advisor is hardened against common prompt-injection attacks (system / user prompt separation, untrusted-content tagging, output validation, pattern pre-screening). Despite these defenses, no AI system is completely immune to adversarial input. If you encounter:
- An AI response that appears to leak data from another user
- A response that ignores its instructions or behaves out of character
- Any unusual or alarming behavior
Please report it to [email protected] immediately. We treat such reports seriously and respond within 24 hours.
7. Your rights regarding AI processing
- Opt-out — you may opt out of AI features by not using the AI Advisor / scanner. Free-tier features that don't depend on AI continue to work.
- Deletion — when you delete your account, we delete your AI conversation history per the Privacy Policy retention schedule.
- Access — your AI conversation history is included in the "Export my data" bundle (Privacy Policy §7).
- Object to AI use — under GDPR Article 22, you have the right not to be subject to solely-automated decisions with legal effects. CarFai's AI features provide information, not legally binding decisions; if you have a specific concern, contact [email protected].
8. Updates
This Disclosure will be updated as our AI use evolves (new features, new models, new third-party AI providers). Material changes will be notified via email and require re-acceptance.
9. Contact
- General AI questions: [email protected]
- Privacy: [email protected]
- Security: [email protected]
Revision history
| Version | Date | Notes |
|---|---|---|
| v1 | 2026-05-20 | Initial publication. |
