AI Safety & Governance
We believe patients deserve to know how the AI analyzing their bills works, what it can and can't do, and how we keep it accountable.
The AI Behind Bill Bodyguard
Bill Bodyguard is powered by Claude, built by Anthropic, one of the world's leading AI safety companies. We chose Claude specifically because Anthropic prioritizes safety, accuracy, and responsible AI development.
Claude analyzes your uploaded bills by examining billing codes, line items, charges, and patterns against established medical billing guidelines. It then identifies potential overcharges and explains its reasoning in plain language.
What Our AI Does
What Our AI Does NOT Do
Transparency means being honest about limitations:
- Does not provide legal advice. Our analysis is informational. Dispute decisions are always yours.
- Does not provide medical advice. We analyze billing, not clinical decisions.
- Does not guarantee outcomes. Hospitals may agree or disagree with flagged items.
- Does not replace professional review. For complex situations, we recommend consulting a medical billing advocate or attorney.
Accuracy & Limitations
We're committed to accuracy, but we're also committed to honesty about what AI can and can't do:
- False positives: The AI may occasionally flag a charge that is actually correct. This is by design, we'd rather flag something for you to verify than miss a real overcharge.
- False negatives: Some overcharges may not be detected, especially for unusual procedures or complex billing scenarios.
- Context gaps: The AI works from the information on your bill. It doesn't have access to your full medical history or the clinical reasoning behind specific codes.
Our recommendation: Always treat AI findings as a starting point. Review flagged items with the hospital's billing department, and consult a professional for large or complex bills.
Bias Mitigation
We take bias seriously. Our AI analysis is informed by established, published billing standards, including:
- CMS (Centers for Medicare & Medicaid Services) guidelines, the federal standard for billing practices
- NCCI (National Correct Coding Initiative) rules, the reference for code bundling
- Typical market rates, based on the AI's training knowledge of common billing practices. The AI does not access live pricing databases or individual hospital chargemasters.
Every bill is analyzed under the same instructions and standards, regardless of hospital, health system, or geographic region.
Data Handling & AI Privacy
When you upload a bill, here's exactly what happens with the AI:
- Your bill content is sent to Anthropic's Claude API via encrypted connection
- Claude analyzes the bill and returns findings
- Results are delivered to you immediately
- Your bill data is not stored permanently by Bill Bodyguard
Anthropic's API data policy states that API inputs are not used to train their models. This means your medical bill data is not being fed into any AI training pipeline. See Anthropic's data policies for details.
Human Oversight
AI identifies potential issues. You make the decisions.
Bill Bodyguard is designed as a tool that empowers you, not one that acts on your behalf. The AI flags potential overcharges and explains its reasoning, but you always decide:
- Whether to dispute a charge
- Which flagged items to pursue
- Whether to send a dispute letter
- Whether to seek additional professional help
We will never automatically file a dispute, contact a hospital, or take action on your behalf without your explicit direction.
Transparency: What We Check
We believe you should know exactly what our AI looks for. Here are the seven categories of billing issues we analyze:
1. Upcoding
When a bill uses a higher-level code than the service provided warrants, for example, billing an office visit as a complex procedure.
2. Unbundling
When services that should be billed together under one code are broken into separate charges, increasing the total cost.
3. Facility Fees
Excessive or inappropriate facility fees, especially for outpatient services that may not warrant hospital-level charges.
4. Supply Markups
Medical supplies and medications billed at significantly above fair-market value.
5. Duplicate Charges
The same service, supply, or medication appearing more than once on a bill.
6. Observation Status Billing
When a private or inpatient room is billed even though the visit appears to have been an observation stay, which is typically billed at a lower rate.
7. General Overcharges
Any other charges that appear priced far above typical rates or inconsistent with standard billing practices.
Continuous Improvement
AI technology evolves rapidly, and so do medical billing practices. We are committed to:
- Updating our analysis instructions as billing guidelines change
- Incorporating user feedback to improve accuracy
- Being transparent about changes and improvements
Our Responsible AI Principles
1. Patient-first: The AI serves patients, not institutions.
2. Transparency: We explain what we do and how.
3. Honesty: We're upfront about limitations and uncertainties.
4. Privacy: Minimal data collection, no permanent storage.
5. Human control: You always make the final decision.
6. Accountability: We stand behind our service and welcome feedback.
Questions or Concerns
We welcome questions about our AI practices. Transparency isn't just a principle, it's how we build trust.
Bill Bodyguard
Austin, TX
Prasanthi@PrasanthiBallada.com
billbodyguard.com