1. How AI Powers Warnvo

Warnvo uses large language model (LLM) technology to analyze the documents and text you submit. When you upload a bill or enter text, our AI reads the content, identifies patterns consistent with overcharging or fraud, and generates a structured analysis and — where applicable — a dispute filing template. The AI does not have access to external databases, real-time pricing information, or your personal financial history unless you explicitly provide it.

2. What the AI Is Good At

• ✓Identifying common billing patterns associated with overcharging (e.g., duplicate line items, inflated service fees, undisclosed charges).
• ✓Recognizing the structure and language of scam communications (phishing emails, smishing texts, fraudulent notices).
• ✓Generating clear, professional dispute filing templates based on the specific charges you identify.
• ✓Reading documents in 67 languages and providing analysis in your preferred language.
• ✓Explaining complex billing terminology in plain English.

3. What the AI Is Not Good At

• ✓Determining with certainty whether a specific charge is fraudulent — it identifies patterns, not facts.
• ✓Accessing real-time pricing data to verify whether a charge is above market rate.
• ✓Understanding highly specialized billing contexts (e.g., complex medical procedures, specialized legal billing) without additional context.
• ✓Guaranteeing that a dispute filing will be successful — outcomes depend on the specific business and applicable law.
• ✓Replacing the judgment of a licensed attorney, financial advisor, or medical billing specialist.

4. AI Accuracy and Error Rates

Like all AI systems, Warnvo's analysis is not 100% accurate. The AI may produce false positives (flagging legitimate charges as suspicious) or false negatives (missing actual overcharges). We continuously improve our models, but you should treat AI analysis as a helpful starting point — not a definitive verdict. Always verify findings before taking action.

5. Human Oversight

Warnvo does not make automated decisions that have legal or financial consequences without your involvement. All AI output is presented to you for your review. You decide whether to act on it, modify it, or discard it. We do not automatically submit dispute filings on your behalf.

6. Data Use and Model Training

We do not use your personal documents to train our AI models without your explicit consent. Aggregate, anonymized usage data (e.g., which types of bills are most commonly scanned) may be used to improve the service. The underlying LLM technology is provided by a third-party AI provider under a contract that prohibits them from using your data for model training.

7. Bias and Fairness

AI models can reflect biases present in their training data. We are aware of this risk and work to ensure our AI does not produce analysis that is unfair or discriminatory. If you believe our AI has produced biased or unfair output, please report it to [email protected].

8. Feedback and Corrections

If the AI produces an analysis you believe is incorrect, you can flag it within the app. Your feedback helps us improve accuracy over time. We review flagged analyses and use them to identify systematic errors in our models.