Overview: Trusting Eve

Last updated: March 9, 2026

Why trust Eve?

Hallucination is a serious concern for any generative AI user—and it poses significant risk to attorneys. The Eve team is dedicated to minimizing hallucination risks through multiple layers of validation and verification built directly into the platform.

Summary:

Eve's approach to trust is based on a "Trust but Verify" framework. While our world-class AI engineering team works continuously to minimize hallucination risks, we recognize that users must be able to easily verify all AI-generated responses.

Our goal is twofold:

  1. Create a platform that minimizes hallucination occurrences through multiple safeguards

  2. Make the "verify" step seamlessly integrated into your workflow

Eve's Verification Features

Eve provides multiple verification tools to help you quickly validate AI-generated content:

Quote Verification (Chat Responses)

When Eve generates responses in Chat, every quote is automatically validated:

  • Green highlights = Exact match found in your documents. Click to jump directly to that location in the source file.

  • Yellow highlights = Similar or paraphrased information found. Click to review the related section and verify accuracy.

  • Red highlights = Information could not be verified in the selected documents. Manually review before using.

Eve runs validation checks after generating each answer, using rules that detect quotes and verify them against your provided documents.

Draft Sources (Drafting Documents)

For documents created with Eve's drafting features (demand letters, complaints, discovery responses), you can verify sources directly within your draft:

  1. Highlight any factual statement in your draft

  2. Click "Sources" to see which case files Eve used

  3. Review exact quotes from the source documents

  4. Jump to the original file to verify in full context

This eliminates the need to copy text and ask Eve separately where information came from—verification is built directly into your drafting workflow.

Inline Sources (All Responses)

Eve automatically cites which documents were relevant for different parts of her answers. Source bubbles appear throughout responses, showing you:

  • Which documents support each section of the response

  • Sections within documents most relevant to that answer

  • The ability to jump directly from answer to source material

Additional Safeguards

Beyond verification tools, Eve employs multiple strategies to prevent hallucinations:

  • Grounding in your documents: Eve bases responses on the specific files you provide, not general internet knowledge

  • Validation framework: Automated rules audit responses for common hallucination patterns

  • Advanced models: Eve uses state-of-the-art language models specifically optimized for legal accuracy

  • Continuous monitoring: Our ML team tracks and addresses reported issues through root cause analysis

Best Practices

To get the most trustworthy results from Eve:

  1. Always verify critical facts using the built-in verification tools before finalizing documents

  2. Select relevant context carefully—the more focused your file selection, the more accurate Eve's responses

  3. Use verification features proactively, not just when something seems wrong

  4. Report any issues you encounter so our team can continue improving accuracy