AI Transparency

AI interaction disclosure

You are interacting with an AI-assisted editing system. AI reviews and edits the submitted text. No human editor reviews the result before it is returned.

Intended purpose

The system removes filler, repetition, clichés, vague abstraction, formulaic AI style, awkward structure, and continuity or logical problems while trying to preserve claims, uncertainty, formatting, and voice. Optional style direction may guide expression but does not authorize invented facts.

Limitations

• AI can introduce or overlook factual and logical errors.
• Preservation warnings check selected literal categories, not truth.
• The slop score is subjective and does not identify whether a human or AI wrote the text.
• A broad rewrite may change tone, emphasis, or meaning.
• Outputs should be reviewed by a person with appropriate subject knowledge.

Output marking and provenance

Every completed result is visibly labelled either “AI-edited with Desloper” or “Substantially rewritten by AI”. Desloper also signs a content hash and edit metadata with Ed25519. The signature can prove that this service emitted that exact revision; it cannot prove that the claims are true or that no later human editing occurred.

EU AI Act position

Our current classification treats Desloper as an assistive standard editing tool where the output does not substantially alter input semantics, which may fall within the Article 50 standard-editing exemption. Substantial rewrites receive a stronger visible label. The classification, semantic safeguards, and need for machine-readable marking are reviewed as the product and applicable guidance change. If the exemption does not apply, compliant machine marking is required for all in-scope outputs rather than through a 200-token runtime block.

Reporting

Report a misleading or harmful output to support@desloper.xyz and include the request ID, but do not email confidential source text unless necessary.