Editorial policy, sources and working method

Which sources take precedence

We work to a fixed order. Higher in this list outranks lower, and where sources contradict each other we say so rather than choosing.

1. Official texts of legislation and regulation. The EU AI Act, the NIS2 Directive and the General Data Protection Regulation, through the published versions, not through third-party summaries.
2. Dutch government and knowledge institutes. The Netherlands Institute for Public Safety (NIPV), the National Coordinator for Counterterrorism and Security (NCTV), the National Cyber Security Centre (NCSC), the Dutch Data Protection Authority, and the regional crisis plans of the safety regions.
3. Academic literature. Including the CERC model from the US Centers for Disease Control and Prevention and the Situational Crisis Communication Theory developed by Coombs.
4. Trade publications and specialist media, always with the name of the publisher and the year.
5. Market research, only with the researcher, the year and the scope stated. We do not include a growth figure without a source.

We do not use anonymous sources. Where something rests on practical experience rather than a publication, we say so explicitly. And where something is still uncertain, for instance because a statutory deadline may still shift, we say that rather than writing it away.

We align with established, recognised methodology rather than an approach of our own devising: the BOB decision model, the GRIP structure, the CERC model and SCCT. That is a deliberate choice. A method that is only ours cannot be tested by you; a method set out in every regional crisis plan can.

How we use AI in writing

The articles on this platform are produced with the help of AI tools, under human control. We think you should know that, and Google also explicitly asks publishers to be open about it.

Concretely, we use AI for organising source material, producing first drafts and checking for unclear phrasing. We do not use it to determine whether something is true. Every factual claim is traced back to the original source by a person. Where an AI tool produces a claim that cannot be found in a source, that claim is removed, even if it sounds plausible.

That is the same working method we advise in our articles: AI speeds up the preparatory work, a person judges and publishes. See also AI hallucinations in crisis communication, which explains why this distinction weighs so heavily for crisis information in particular.

How often articles are reviewed

Under every article you will find when it was published and when it was last reviewed. That second date changes even when nothing has been amended: it means someone laid the content against the sources again on that day.

Type of article	Review frequency	Why
Regulation (EU AI Act, NIS2, GDPR)	Quarterly	Deadlines and obligations are still shifting
Risks (deepfakes, hallucinations)	Every six months	The threat picture changes quickly
Practice (applications, tools)	Every six months	What is technically possible keeps moving
Foundations (GRIP, BOB, CERC)	Annually	The methodology is stable

Independence

This platform is editorially independent of the publisher's product marketing. In practice that means three things.

We do not write an article to sell a product. Topics are chosen on the basis of what crisis teams actually ask about, not on the basis of which feature we happen to offer. Every article has to stand on its own, even if you never click through.

We do not name competing products, in either direction. Not to promote them and not to disparage them. Where it is useful to discuss a category of solutions, we do so generically: "standalone AI text generators" rather than a brand name. A comparison in which the publisher competes is not a fair comparison, and we are not going to imitate one.

We mark references to our own services. Where we refer to CrisisRadar or CrisisMaatje, we state that it is a service of the publisher. Those references sit in separate blocks, not hidden in running text as if they were neutral recommendations.

Funding and ownership

AI Crisis Communication is funded entirely by the publisher. There are no advertisers, no sponsored articles, no paid listings and no affiliate fees. We do not sell links and we do not accept payment for including or omitting a topic.

The publisher builds software for crisis management and crisis communication and has a commercial interest in crisis teams being well informed about AI. We do not conceal that interest; it is stated here and on About this site. What we do promise is that this interest does not steer the content, and that you can hold us to that.

Corrections

Is something wrong, is a source out of date, or are we missing something important? Let us know at redactie@aicrisiscommunicatie.nl or through the contact page.

For a substantive correction we amend the article, update the review date and record the change in the public corrections log. We do not remove errors quietly. That is a deliberate choice: a platform writing about reliable information during a crisis cannot allow itself to make its own mistakes invisible.

What this platform is not

This is not legal advice, not a regulatory ruling, and not a substitute for a crisis communication adviser, lawyer or data protection officer for your own situation. Where an article touches on obligations applying to your organisation, have the application tested by someone who carries responsibility for it. That is also stated in the articles themselves, at the points where it matters.