AI Use Policy

This is Ennui Nation’s policy on the use of artificial intelligence. It sets out where AI is used in the work, where it isn’t, and who answers for what gets published. It applies to everything here: the reporting, the writing and editing, and the running of the publication. Where anything shorter elsewhere on the site touches on the same subject, this is the version that stands.

The principle

AI is a tool I use to work faster. It isn’t a journalist or an editor, and it doesn’t get to decide anything. It has no authority over what Ennui Nation covers or publishes. A person does the judgement, and a person carries the responsibility. That person is me, the editor.

Where AI is used: editorial work

For the reporting, writing and editing on this site, I use AI for the early and practical parts of the process:

• Getting a rough first draft down, which I then rewrite in my own words.
• Working out how a piece might be structured or ordered.
• Speeding up early background reading, so I’m not starting from nothing.
• Tidying notes, summarising long documents I already hold, and catching typos and clumsy phrasing.

None of this is the journalism itself. It’s the scaffolding I build on and then take down.

Where AI is used: predictions and forecasting

Some of Ennui Nation’s work, particularly around elections, involves forecasts: probability-based calls about what is likely to happen, rather than conventional reporting. AI does much of the number-crunching. I set the method: what data goes in (past results, polling, demographics and coverage from across the political spectrum), the rules the model has to follow, and how every call is scored. AI then works through the data within those rules and produces the probabilities, and nothing is published until I’ve signed it off.

Every forecast is locked before the polls close and published with its reasoning and its likely range, not as a bare number presented as certain. Once the result is known, each forecast is scored against it, alongside other major forecasters, and the lessons go back into the method. A published forecast is never quietly edited afterwards. This work is kept separate from editorial judgement: it doesn’t decide which stories run or how they’re framed.

Where AI is never used

Some parts of the work AI has no place in, and these are firm:

• What gets covered, and how a story is framed, is never its call.
• The finished journalism is in my own words. Anything published has been rewritten and shaped by me, not left as a first draft.
• Facts get checked independently, not taken on the tool’s say-so (see below).
• Quotes, sources and evidence all come from real material. Nothing invented by a tool is ever published as fact.
• Legal, ethical and editorial judgement is mine alone.
• Sources are never contacted, identified or profiled through it.
• AI-generated images, audio or video are never passed off as real. Where one is used to illustrate something, it’s labelled clearly.

Checking facts

Every factual claim is checked against primary sources: original documents, records, direct statements, named evidence. AI is not a source and is never treated as one. If a tool suggests a fact, a figure or a quote, it doesn’t go in until it’s been confirmed independently. Anything that can’t be stood up doesn’t run.

Sources and confidentiality

Confidential material is kept away from general-purpose AI tools entirely. That covers unpublished source material, anything touching the identity of a source, investigative notes, and correspondence given in confidence. This work is handled within the Proton ecosystem of encrypted services and isn’t shared with third parties. Protecting a source comes before convenience, always.

Data and privacy

I take care over what goes into any AI tool. Personal data, sensitive information and confidential documents are not fed into general-purpose services. Where these tools are used for ordinary drafting or research, what’s shared is kept to a minimum, and anything put in is treated as though it could be retained. Readers’ personal data is covered in the Privacy Policy, and AI isn’t used to profile or track anyone who visits the site.

Copyright and where the words come from

Work is never drawn from an AI tool’s memory or training data. Anything current or time-sensitive gets checked fresh, against several original sources, right before it runs. Nothing current is ever assumed from something checked days or weeks ago. Once a fact is properly checked and verified, it isn’t re-proven from scratch every time it comes up again, but that’s not the same as treating it as good forever: nothing stale or unconfirmed goes into anything current. From there I write the article, and rewrite it until it’s right, before it goes back through fact-checking and verification. Nothing here is copied or lifted from someone else’s writing. If a passage ever echoes someone else’s work too closely, get in touch and I’ll look into it.

Who is accountable

Every editorial decision and every published word at Ennui Nation is made and authorised by a person: me, the editor. If something is wrong, it’s my mistake, and mine to put right. The corrections process and the complaints process both come straight to me. Neither is handled by an automated system.

If AI gets something wrong

If an error traces back to something AI produced, I don’t hide behind the tool. I made the call to publish it, so it goes through the same corrections process as anything else. AI doesn’t carry the blame for a decision I signed off.

Telling you when AI is involved

If AI has played any part in an article beyond the ordinary drafting and research described here, I’ll say so on the article itself. The position here is openness: you’re entitled to know how the journalism you’re reading was made.

AI-based services for readers

Ennui Nation is also developing separate AI services for readers that give general information and guidance, not advice. They are distinct from its journalism. Those services process different information in a different way, and where and when they launch publicly, they will carry their own dedicated policy rather than being folded into this one.

Keeping this up to date

I keep this policy current as the tools change and as my working methods change. If I start using AI differently, this page is updated before anything else.