Our policy on artificial intelligence (AI)
NK News, NK Pro and Korea Pro products are written, reported and edited by people. Every article carries the byline of a named human author, and every article is read and cleared by a human editor before publication. AI tools may help our journalists work more efficiently in areas noted below, but the reporting, sources, analysis and conclusions in every article are our journalists’ own.
We also use AI tools in parts of our production process, described in this policy, and in our business operations, described in our Privacy Policy. Subscribers are entitled to know exactly where we use AI, so this page sets out what we use these tools for, what we will not use them for and how we protect our sources.
What we mean by AI
In this policy, “AI tools” means generative AI services, such as large language models and AI transcription, translation and image tools, that process content we send them. This includes AI features built into other software we use. Staff may use only AI tools approved by our technology team, through company-managed accounts.
Where we use AI
Language and editing. Several of our staff are non-native speakers of English and use AI tools to correct grammar and improve phrasing in their own drafts. Our journalists may also use AI tools to help draft, restructure, condense or edit their own work. The bylined author is responsible for every word and every claim, and an editor reviews every article before publication.
Transcription. We use AI transcription services for interviews, subject to the source protection rules below. Every quotation is checked against the original recording before publication.
First-pass translation. We get first-pass translation assistance when monitoring primary sources in Korean, Chinese, Russian, Japanese and other languages that are part of routine monitoring relevant to our beats. Machine translation is often the first step, but it is never the last: Any quotation or claim that reaches publication has been checked against the original-language source by a person who is native in that language. AI translation is also being provisioned for non-English speakers to access our website, but English content will continue to be the authoritative copy.
Data processing. We work with large document sets, trade statistics, shipping records and satellite imagery. We use AI and other automated tools to search, sort, extract and flag patterns in this material. Any finding that appears in an article is verified against the underlying records before publication.
Charts and graphics. We use software with AI features to produce charts, maps and page graphics.
Internal summarization. Our staff use AI tools to summarize long documents for their own reading. Those summaries are working notes, not a substitute for reading the source, and are not published. The exception is our Month in Review, parts of which are AI-assisted summaries of our own published reporting; those parts are labeled as such.
Background research. Our staff use the internally-developed Intelligence Suite tool, which uses the NK News, NK Pro, Korea Pro and KCNA Watch corpus, as well as company-designed and managed connectors to external AI tools, to quickly aggregate what our outlet reported on in the past for background and context explanation.
SEO, newsletter optimization and social media. We may use AI tools to suggest headlines, newsletter blurb drafts, social media posts and search-optimization text such as meta descriptions. These are suggestions only, however, and a journalist or editor chooses, rewrites as needed and approves the final wording, and checks it for accuracy against the article.
Where we do not use AI
We do not use AI to do our reporting or analysis for us. Authors may use AI to test their analysis, for example by asking what counterarguments exist, but every analysis we publish is one the author has produced, evaluated and stands behind.
We do not use AI to generate quotations, sources, citations or attributed statements. Every quotation in our copy comes from a real exchange, document or broadcast.
We do not treat AI output as a source. An AI system’s claim about North Korea is not evidence of anything and does not enter our reporting.
We do not publish AI-generated or AI-altered images as documentary photography or any photo-realistic graphics. Where an illustration is generated with AI, we label it as such in the caption.
We do not create synthetic bylines, contributors or personas.
Source protection
Our reporting depends on people who take real risks to speak to us.
We never enter source-identifying information, or material from sources whose safety could be at risk, into AI tools, including transcription services. On-record interviews and public material may be processed with approved AI tools.
Background or off-record material from sources who are not at risk may be processed only with approved tools whose terms bar training on our data and limit how long it is kept, with identifying details removed where feasible.
Handling of sensitive material is governed by our source protection practices, which apply to AI tools as they do to any other third party.
All KRG staff are required to agree to the company’s technology policy, which outlines AI best practices and data privacy guidelines.
How we collect, use and share the personal data of our subscribers, customers and others, including where Artificial Intelligence Service Providers process it, is set out in our Privacy Policy.
Accountability
Every article passes through our editing desk before publication. Our editors are accountable for what we publish, and an error introduced by a tool is our error, not the tool’s. Where we get something wrong, we correct it and say so, under the same corrections policy that applies to all our work. On rare occasions where AI-generated text makes up a substantial part of published content without being rewritten by a journalist, we say so on the page.
Our contributor agreements hold freelance and guest contributors to this policy. Contributors must disclose any use of AI tools in preparing work for us.
This is a working policy in a field that is changing quickly. We review it periodically and will update this page when our practices change. Questions about it can be sent to Ethan Jewell, Director of Technology ([email protected]).
Last updated: October 22, 2026
