AI disclosure

1What AI writes

RiftCommand uses AI that writes text (a large language model). It writes:

• every scene and line of dialogue in your missions, for your crew and for the contacts you meet;
• each mission’s situation, people and possible endings, drafted before the mission starts;
• your officers’ crew files when they join: names, backgrounds and ways of speaking;
• the orders suggested to you during a mission;
• the Bridge Log, the epilogue and the words on a memorial at the end of a mission.

Nobody writes or reviews these lines before you read them. Each one is AI-generated for your orders, so no two missions read the same.

2What people wrote

People at RiftCommand wrote the world (its species, factions, history and technology), the rules of the game, the instructions that shape what the AI writes, and the text that isn’t story: buttons, menus, messages and these pages.

3How an order is decided

The AI tells the story; it doesn’t decide it. Each order goes through these steps:

1. The game’s code sets aside an order it can’t read at all; your crew asks you to try again, and it doesn’t count as an order.
2. A smaller AI reads your order and sorts it into one of five kinds of action (investigate, confront, embrace, ignore or negotiate), and notes whether it is a question and how specific it is.
3. The code works out what happens, by fixed rules: how well the action goes, what it costs, and how your ship, your crew’s trust and your resources change. A question changes nothing.
4. The code tells the AI what happened, and the AI writes the scene.
5. The code applies the changes and decides whether the story moves on to its next act.

When a decision is due, the code matches your order to one of the options shown by comparing their words. Which ending a mission reaches is chosen by the code, from the decisions you made. The AI can’t change any number in the game or pick an outcome.

4What can go wrong

• It can get the story wrong: forget something that happened, contradict an earlier scene, or mix up who said what.
• It can repeat itself and use the same phrases again and again.
• It can misread an order. If your crew answers something you didn’t ask, give the order again in other words.
• It can write something unsuitable or offensive despite our instructions. We don’t run a separate filter over the story: we rely on our instructions and on the safety measures of the companies whose AI we use.
• Missions deal with hard choices, violence and loss by design.
• Nothing it writes is fact or advice. The characters and events are fictional; any likeness to real people is unintended.

5The AI we use

The story is written by [the AI chosen after testing, and who makes it], reached through OpenRouter, a service that passes each request to a company running that AI. We may change the AI we use, and we update this page when we do.

The AI receives your order and what it needs to continue the story, never your email address or payment details. Each request is first offered to free AI services, whose providers may keep what they receive and train AI on it, so keep personal data out of your orders. When no free service is available it goes to paid providers, which we set OpenRouter to choose only among those that, by their own policies, don’t keep requests after answering them or train on them. We don’t use what you type to train AI ourselves. The privacy notice has the details.

• Privacy notice

6Voices and pictures

• Crew voices are synthetic speech, made on your device by your browser’s voices or, if you get it, the voice pack (Kokoro, an openly licensed speech model). No actor recorded them, and the words are always on screen too.
• Where the game shows painted portraits, places or ships, they were made with an AI image tool and chosen by us. The line-drawn stand-ins you may see instead are drawn by the game’s code.

7Reporting a line

If a line is harmful, offensive or badly wrong, tell us at [Contact email]. Include the mission (its sector and the date), the order, and the line, quoted or as a screenshot. We read every report and use them to improve the instructions the AI follows. We answer within [reply time, for example 5 working days].

If the line involves something illegal or a risk to someone’s safety, say so at the top of your message and we will deal with it first.

8Where we tell you

We say plainly that the story is AI-generated: on the home page, in the answers to common questions, at the foot of every public page and at the foot of the bridge in the game.

This page sets out what Article 50 of the EU Artificial Intelligence Act (Regulation (EU) 2024/1689) asks us to make clear: that you are dealing with AI, and which content it wrote.