AI can assist with drafting, code generation, organization, testing ideas and research workflows, but it does not replace source verification, deterministic calculation tests, jurisdiction evidence or specialist review. Publication controls are designed around the evidence and the user risk, not around whether a draft was human- or machine-assisted.
What matters
- AI-generated text is not itself evidence for a technical claim.
- Primary sources and controlled data remain the reference point for factual and technical assertions.
- Formula runtimes are tested independently of prose generation.
- Professional review requirements do not disappear because an automated system produces a confident answer.
How to use this page
This policy describes the controls used for the current public site. It is written to match implemented behavior: the launch excludes unresolved specialist-review process pages, keeps source and review states separate, and does not claim professional review where no verified reviewer exists.
Automation can assist production; it cannot create evidence
AI-assisted drafting or structuring can improve workflow, but it does not make an unsupported claim true. Sources, calculations, publication decisions and review states need independent controls.
No fabricated evidence
An automated system should never invent citations, reviewers, measurements or project facts.
Method stays testable
Formula behavior and numerical outputs should be verified through deterministic tests rather than trusted because a model produced them.
Human judgment where needed
Editorial, legal, specialist and publication decisions remain distinct from text generation.