POLICY AND TRUST

Zero-AI and 100% Originality Policy

Human-created work, provider-verified before release, with a checksum and audit trail tied to the exact final file.

Effective 24 July 2026 · Version 2.0

The standard

Every written deliverable must be created for the approved request by an accountable human specialist. Generative AI must not write, paraphrase, expand, or disguise final client-facing text.

100% original means no knowingly copied, purchased, recycled, or unattributed material. Correct quotations, references, common phrases, formulas, and technical terms can still create legitimate similarity.

Mandatory API verification

  • Copyleaks Authenticity API is the default provider.
  • Turnitin may be used only with a valid Similarity, Originality, or SimCheck license and organization-specific API credentials.
  • A screenshot, manually typed score, browser extension, or unverified report cannot unlock completion.
  • The system records the provider, scan ID, SHA-256 checksum, filename, result, threshold, and pass or fail decision.

Release rule

A file passes only after a successful provider scan, similarity at or below the configured threshold, no suspected-AI alert, and human quality review. The default similarity threshold is 20%.

A low score is not permission to copy, and a high score can include legitimate cited material. Human review remains decisive, but a provider failure cannot be represented as a pass.

Verification privacy

Provider transfers use TLS. Copyleaks shared-database indexing is disabled and provider expiration is limited. Signal Academic stores a content checksum and result metadata rather than a second persistent copy solely for scanning.

Failed results

A failed file returns to quality review, must be corrected, and must be scanned again. Staff cannot bypass the completed-state gate. Apparent false positives require documented human review and may be rerun through another licensed provider.

Prohibited conduct

  • Generative AI authorship in final client-facing text.
  • Plagiarism, patchwriting, recycling, or undisclosed reuse.
  • Fabricated citations, data, quotations, cases, or results.
  • Hidden text, character substitution, image conversion, or other detection evasion.
  • Altered, suppressed, or fabricated provider reports.
  • Scanning a different file from the file delivered.

Remedy

A verified breach requires replacement or correction at no charge. If it cannot be cured within the agreed period, the unearned amount is refundable under the Money-Back and Revision Guarantee. The specialist may be suspended pending review.