Below are answers to common questions about how our AI features process your data, protect sensitive information, and help your firm stay compliant.

Common Concerns:

1. Are my documents or customer data used to train AI?

No. Your documents, prompts, inputs, and generated summaries are never used to train, fine-tune, or improve our AI models. They are also never used to train any third-party foundation models. Your data remains entirely yours.

2. How do you handle privacy and confidentiality?

All data processed by our AI is subject to the same security measures as the rest of the platform. Our AI infrastructure is deployed via Amazon Web Services, so all processing happens within our closed, enterprise-grade cloud environment. Your data is never sent to public AI platforms such as the public version of ChatGPT or standard web-based AI interfaces.

3. How do you treat confidential or highly sensitive information (PII/PHI)?

Because we service the legal industry, our system is designed to treat all input data as strictly confidential.

• Encryption. All text and documents processed by the AI are encrypted in transit and at rest.
• Closed processing. When a document is analyzed, it is processed entirely within your secure tenant. The AI reads the document only to generate your requested output — such as a document summary — and does not retain the information outside of your specific workspace.

4. Where is data stored, and for how long?

• Document storage. Your original documents remain in your native document storage environment, governed by your existing retention policies.
• AI outputs. Generated outputs — such as document summaries — are saved directly to your firm’s secure tenant so you can access them without re-processing the document. These outputs are retained only for as long as you keep the underlying document on the platform. If you delete the document or make changes to it, the AI summary is deleted as well.

5. Is anything cached during AI processing?

To provide a fast experience, we temporarily hold the text of a document in memory just long enough for the AI to read it and generate the output — typically a matter of seconds. Once the output is generated and delivered to your tenant, this temporary data is immediately discarded.

We do not maintain ongoing caches of your document text on our AI processing servers. The final generated summary is saved securely in your native storage alongside the original file to optimize future load times.

6. How is conversation history or prompt history handled?

For the document summary feature, there is no ongoing conversation or conversation history. The AI performs a one-time analysis and generates a static output.

For any interactive AI features, your input history is securely isolated to your firm’s tenant and user account. It is retained only to support a continuous experience and is never pooled with other customers or used for model training.

7. What AI model powers this feature?

We use Anthropic’s Claude Haiku, hosted securely through AWS Bedrock. This model was selected for its speed, reliability, and precision in extracting factual data from complex legal text, all within a secure enterprise infrastructure.

8. Will the AI give me legal advice or hallucinate facts?

Our AI is designed as a read-only, non-advisory assistant.

• No legal advice. The AI is programmed to describe what is in the document using neutral language. It will not analyze precedents, offer legal reasoning, flag risks, or make recommendations.
• Preventing hallucinations. The AI operates under strict extraction rules. Features such as Key Facts and Outline will only display information explicitly written in the document. If a document lacks clear headings or specific facts, the AI will leave those sections blank rather than attempt to fill them in.

9. Can our firm opt out of AI features?

Yes. If your firm’s internal policies or client agreements require you to abstain from AI usage, your firm’s administrator can contact our Support team to have the AI Document Summary feature completely disabled for your tenant.

Note: To request this, contact support@mangopractice.com and reference your firm’s tenant name and the feature you would like disabled.

Learn More

The following knowledge base articles contain additional information relating to AI:

• AI Document Summary