How We Use AI
🔗Human Responsibility
Work on the UPDV began in 2003, and the bulk of its translation work was completed before AI assistance was introduced around 2023. Every final change is approved by the editor. AI output, specialist recommendations, or agreement among reviewers do not determine the wording of the biblical text. The UPDV should not be considered an AI-generated Bible.
🔗What AI Assists With
Since 2023, AI has assisted with processing the existing backlog, retrieving background research for new questions, investigating textual issues, proofreading for errors and archaic English, checking consistency, producing media, and handling technical publishing and maintenance work. In many cases, it has also helped turn earlier notes and decisions into fuller explanations of the evidence and reasoning behind them. AI assistance was also instrumental in making possible the UPDV Bible Assistant Chat. Grounded in the translation and the published articles and topic pages, the assistant allows readers to ask questions that draw together findings from across the project.
🔗Why We Use AI
AI assistance is a necessary tradeoff for operating at this scale and processing the kinds of research the project undertakes. More than 3,000 research questions, reviews, corrections, and technical tasks have been tracked. The work ranges from comparing original-language and manuscript evidence to checking citations and keeping the biblical text, footnotes, articles, and topic pages consistent. Even a small wording decision can require reviewing and synchronizing many related passages and supporting materials. This assistance makes an otherwise impractical amount of research and technical work manageable while allowing the translation to remain freely available in the public domain.
🔗Evidence, Review, and Authority
AI output is not itself textual, lexical, grammatical, or historical evidence. Research claims must rest on retrieved source documents and scholarly literature. Questions are divided into focused tasks and may be reviewed by specialized AI research roles for source languages, textual evidence, literature, or citations. These roles assist the editor; they do not replace the editor's judgment. The review process checks quotations and citations against retrieved source text and records corrections when errors are found.
🔗Limits
AI can make mistakes. It can fabricate claims or citations, retrieve only part of a source, misread internal references, reflect model biases, or give results that vary between runs. No generated claim is accepted merely because a model stated it or multiple reviewers agreed. When an error is found, the affected Bible text, article, footnote, topic, or technical surface is corrected. Readers who discover an irregularity are encouraged to use the “Contact our help desk” link at the bottom of the page.
🔗External Services and Data
The project uses external AI-model and managed-retrieval providers. Depending on the task, material sent to these services may include a reader's question and recent chat history, search queries derived from the question, relevant UPDV Bible passages, published articles and topic pages, and excerpts from research sources. Public UPDV Bible text and published explanatory material may also be stored in externally managed search indexes. The website keeps local operational logs containing a visitor's IP address and up to the first 500 characters of the submitted message.
The Bible Assistant does not require private or sensitive information. Because questions and recent chat history are sent to external providers, readers should not include confidential, private, or sensitive information in their questions.
🔗Keeping This Statement Current
This statement describes the UPDV's use of AI as of September 2026. It will be updated when the project's AI-assisted workflow changes materially.