AIUse&Transparency

AxonCraft is built on AI. Reading briefs, writing and developing ideas, writing notes on them and drawing case boards are all done with AI systems. Presentations are built from your saved idea without an AI model.

The EU AI Act requires providers of AI systems to be clear about where AI is involved, and to mark content that AI has generated. We think that is the right standard regardless of the law, because you cannot judge creative work well if you do not know how it was made.

This page is a summary written in plain language. It is not a complete statement of the Regulation, and it is not legal advice. Where this page and the Regulation differ, the Regulation governs.

AxonCraft uses AI models from three providers: Anthropic (Claude), OpenAI (GPT) and Google (Gemini). This is how a project runs, step by step.

• Your brief. You type a brief, or upload documents and images. When you upload, a Claude model reads the files, including the text in images, and proposes a brief that you check and edit before saving. When you save a brief, a Claude model reads it and picks the strategic angles to start from.
• Strategy. A Claude model writes four strategic starting points for each round. From a project's second round, it also sees what earlier rounds of that project explored, so it does not repeat them.
• Ideas. Four AI writers each write two ideas, so every round gives you eight: two Claude models, one GPT model and one Gemini model. You can ask for eight more.
• You pick. You choose which ideas to take further. No model makes that choice for you.
• Develop. The model family that wrote an idea develops it into a fuller case and execution. The two other model families then write craft notes on it. They are notes, never a verdict, a score or a rewrite.
• Shape and Refine. A Claude model rewrites an idea or its execution to answer your note.
• Presentation. You can export an idea as a presentation. It is built from your saved idea without an AI model.
• Case boards (optional). If you ask for a case board, a Claude model writes the instructions and an image model from Google (Gemini) draws it.
• Small helpers. A Claude model writes the short "what changed" line after a long brief edit. Where enabled, it also rewords an idea's title and description to remove stock phrases; it cannot change the idea itself.

• No AI ranks, scores, filters or picks among your ideas. You see all eight ideas from each round. A check in our code, not a model, makes sure each idea has its required parts. If a writer's answer fails that check twice, the round stops and tells you; it does not quietly show you fewer ideas.
• AI does choose the strategic starting points the writers work from, and the craft notes are AI opinions. They are notes, never a verdict or a winner.
• Search data and presentations use no AI model. When you save a brief, we look up public search interest for up to three names your brief repeats. The results are shown as search data, and no model turns them into insights. Presentations are built directly from your saved idea.
• You decide which ideas to shape, develop, make final, present and use.

The app shows a visible AI label in these places. Each idea's own page shows AI drafted, or AI drafted · human edited once you, or AxonCraft at your request, change it. Project cards on the Projects page show AI-generated. Board images are labelled AI-generated image.

Labels are not a setting. There is no plan, preference or upgrade that removes them.

Presentations carry AI drafted and the line "Generated with AI. Review before use.", so the notice travels with the file when it leaves the app.

We do not remove the C2PA provenance data that image providers embed in the images they return. We check each generated image and write a warning to our server log when that data is missing.

Reliable machine-readable marking of generated text is still an unsettled technical area, and standards are developing. Where we cannot mark content in a machine-readable way, we rely on visible labels instead.

AxonCraft has no feature that trains or fine-tunes an AI model with your briefs or ideas.

The conversational assistants in AxonCraft are AI systems, not people. No human reads or answers your messages in these surfaces.

We tell you this in the interface, before you interact, on every surface where you type instructions or questions to a model. That notice is not dismissible and does not depend on your plan.

AI assistants get things wrong, including confidently. Treat what they tell you as a suggestion to check, not as a fact to rely on.

AxonCraft does not operate AI in the areas the AI Act treats as high-risk. We do not use AI for recruitment or worker management, credit scoring, education or exam assessment, biometric identification or categorisation, or law enforcement.

Our output is a starting point for human judgement. It is not a decision about any person, and it should not be used as one.

We re-check this classification when a feature's function changes materially, or when the kind of people using it changes. A tool that starts inferring things about people rather than about ideas is a different tool, and we treat it as one.

If you publish, broadcast or otherwise release content you made with AxonCraft, you may have your own transparency duties as a deployer of an AI system. Two cases matter most.

• Content presented as informing the public on matters of public interest normally has to be disclosed as artificially generated or manipulated.
• Content that depicts a real, identifiable person doing or saying something, where that depiction is generated or manipulated, normally has to be disclosed too.

Keep our labelling in place, or make your own disclosure. Removing a label does not remove the duty, and it is your duty rather than ours once you publish.

Some uses of AI are banned outright, and we do not build them. AxonCraft does not use:

• manipulative or deceptive techniques designed to distort behaviour
• social scoring of people
• biometric categorisation of people
• emotion inference in workplace or education settings

We also expect you not to use AxonCraft to build these things.

Questions about how we use AI, or about anything on this page, can go to legal@axoncraft.ai.