IntelliNews produces original reporting on more than 130 emerging and frontier markets, written by correspondents in the countries we cover. Subscribers pay us for judgement they can act on, and that judgement comes from people.

AI tools are now part of how most publishers work. Readers are entitled to know where those tools sit in our process, where they do not, and who remains accountable for what appears under our name. This policy sets that out. It also sets out the terms on which other parties may use IntelliNews content in AI systems.

It applies across the business: editorial, data products, marketing, commercial and technology.

A named human is accountable for everything we publish. Every article carries a human author or an identified editorial team, and an editor signs it off. Nobody at IntelliNews can point at a tool when something goes wrong.

AI assists, it does not report. Reporting means talking to people, reading documents, being in the country and understanding the politics behind the numbers. No tool does that for us.

We do not publish synthetic content without saying so. If AI has generated something a reader could reasonably mistake for reporting, photography or a real quote, we label it or we do not run it.

We use AI tools in limited, supervised ways, including:

• Translation. We cover markets in dozens of languages. Machine translation gives our journalists a first pass at source documents, filings, local media and official statements. Translated material is read in context by the journalist reporting the story, and checked against the original wherever we have a reader for that language. Where we cannot verify exact wording, we report the substance rather than quoting directly.
• Transcription. Interview and briefing recordings, transcribed for the journalist's own use. The journalist checks any quote against the recording before publication.
• Research support. Summarising long documents such as central bank reports, IMF filings, tenders and legislation, so a reporter can find what matters faster. The underlying document is always read before we report on it.
• Data processing. Cleaning, structuring and cross-checking large datasets in our data products, and flagging outliers for human review.
• Sub-editing support. Suggesting headlines, standfirsts and subheadings, which editors then accept, rewrite or reject.
• Marketing and commercial operations. Drafting promotional copy, campaign assistance and internal analysis. Marketing copy is reviewed by a person before it goes out.
• Drafting support. AI may produce a first draft of a section of an article. A journalist then edits it, verifies it against sources and takes authorship of the finished piece. Nothing reaches a reader without that step.
• Illustration. We sometimes use AI to generate abstract or conceptual images. These are labelled as AI-generated.
• Product and software development. Coding assistance in building and maintaining our platforms.

• We do not publish AI-drafted text that no journalist has read, edited and taken responsibility for.
• We do not use AI to generate sources, quotes, interviews or attributed comment.
• We do not use AI to generate or alter images depicting real people, places or events. Our AI-generated imagery is abstract or conceptual, and labelled.
• We do not let a tool make the editorial call on what is true, what is significant, or what is fair to publish about a person or a company.
• We do not enter confidential source material, unpublished investigative work or personal data about sources into third-party AI tools.
• We do not use AI to produce automated ratings, rankings or awards outcomes without human assessment.

AI tools make confident mistakes. That is precisely why we keep verification with people. Our correction process is unchanged: if we get something wrong, we fix it, and we say we have fixed it. Errors made with the assistance of a tool are still our errors.

To flag an inaccuracy, contact [email protected].

We handle personal data under the GDPR and our Privacy Policy. Staff and contributors work under confidentiality terms that restrict what may be entered into external AI services, including a prohibition on uploading our archive or any substantial part of it to third-party tools. Subscriber data is not used to train third-party AI models.

Some IntelliNews analysis is produced with the assistance of IntelliNews Lambda, our own analytical tool.

Lambda is a machine learning system built on two decades of IntelliNews daily reporting and the hard data underlying it, covering around 100 countries across Eurasia, Emerging Europe, Central Asia, Latin America, the Middle East, Africa and Southeast Asia. It combines granular reporting with quantitative analysis to produce assessments that can be expressed numerically as well as descriptively, including measures of alignment and probabilistic outcomes.

Lambda is not a large language model, and we have not trained a language model on our archive. It does not write copy in place of a journalist. It produces analysis that our journalists and editors then interpret, verify and take responsibility for publishing.

Where Lambda has contributed to a piece, we say so on that piece.

Our website search is an AI system built by IntelliNews using a third-party model. It interprets questions asked in ordinary language and returns the IntelliNews articles that answer them.

It retrieves our journalism. It does not write news, and it does not generate articles or summaries presented as reporting. Subscription access is unchanged: articles behind our paywall stay behind it. The search is identified as AI-powered where you use it.

We use AI services from established commercial providers under business terms that exclude our inputs from being used to train their models, wherever such terms are available. We review new tools before adoption and assess them for accuracy, security and data handling.

IntelliNews content, including articles, headlines, photographs, graphics, databases and data products, is protected by copyright and database rights owned by IntelliNews Media OÜ or its licensors.

IntelliNews expressly reserves all rights in its content for the purposes of text and data mining, including the reservation permitted under Article 4(3) of Directive (EU) 2019/790 on copyright in the Digital Single Market and equivalent provisions in other jurisdictions.

Without a written licence from us, you may not:

• Use IntelliNews content to train, fine-tune, ground or evaluate any AI or machine learning model.
• Scrape, crawl, harvest or systematically download our content, whether by automated means or otherwise.
• Use IntelliNews content in retrieval-augmented generation, embeddings, vector databases or any similar system.
• Reproduce our content, in whole or in substantial part, in the output of an AI system.
• Circumvent our paywall, authentication or access controls for any of the above.

This reservation applies to our content however it was obtained, including through our website, RSS feeds, email newsletters, syndication partners, aggregators, third-party archives and previously collected datasets.

We enforce these rights.

We are open for business on this. IntelliNews licenses its content and data to terminals, aggregators, research platforms and AI developers on commercial terms, and we would rather agree a licence than send a legal letter.

For licensing enquiries, contact [email protected].

Our machine-readable preferences are published in our robots.txt file. Crawlers must respect it. Ignoring it does not create any permission, and permission is not implied by our content being publicly reachable.

Staff and contributors who work with these tools receive guidance and training on what the tools can and cannot do, and we keep a record of it.

This policy is owned by IntelliNews management and reviewed at least annually, or sooner if our practice, the tools available or the applicable law change materially. Where the EU AI Act imposes transparency obligations relevant to our use of these systems, we comply with them.