AI policy

Meaning first. Style second. Human judgment last.

SayMorph uses AI to transform delivery, not to establish truth. Our product and prompt rules are designed around that distinction.

Last updated August 16, 2026

Three deliberate versions

Safe, Balanced, and Bold are generated together so users can compare the distance from the source. Stronger style is never permission to alter facts or invent context.

Hard factual invariants

Names, organizations, products, amounts, percentages, dates, times, IDs, URLs, email addresses, requests, refusals, deadlines, commitments, and negative relationships are checked server-side. A result that fails those checks is rejected rather than silently presented as successful.

Prompt-injection boundary

Submitted text is treated as untrusted source material. Commands, roles, markup, links, and phrases such as “ignore previous instructions” inside that text are content to transform—not instructions for the model or server.

Style and identity boundaries

SayMorph supports broad tone, era, genre, and original-character traits. It does not support imitation of celebrities, living public figures, brands, protected characters, or a living writer’s distinctive style.

Human review

Always verify important output. Legal, medical, financial, contractual, employment, safety, and public statements require appropriate human or professional review.