Use of AI at InfiArtt

1. Our principles

We use artificial intelligence (AI) to take on repetitive, time-consuming work, such as checking spelling, transcribing recordings, or translating articles. This lets our team focus on the things that need human judgement.

We follow three principles when using AI:

• People stay in charge. AI makes suggestions; editors, organisers, and authors make the decisions.
• Openness. You have a right to know when something was made or assisted by AI.
• Only what is needed. AI is given only the content its task requires, nothing more.

---

2. Meet Arunika

Arunika is InfiArtt's AI editorial assistant. Arunika has its own account, so its notes during article review always appear under Arunika's name, never under a team member's. That way you can always tell AI suggestions apart from an editor's notes.

What Arunika does:

• reads articles submitted for publication and leaves notes on spelling and grammar;
• assesses whether image alternative text describes what the image shows;
• flags statements that should be backed by a source;
• suggests a summary and title;
• suggests a cover illustration for articles that do not have one; and
• writes draft translations of Knowledge Base articles under its own name, crediting the original author. A person checks each draft before it is published.

What Arunika does not do:

• publish, decline, or change the content of an article;
• access account data such as email addresses, phone numbers, or passwords; or
• read private messages, support tickets, or WhatsApp conversations.

Editors and authors can accept, ignore, or disagree with Arunika's notes. Because AI can be wrong, we treat its notes as input to consider, not as a final verdict.

---

3. Where AI is used

• Article review. AI reads the text and images in submitted articles and leaves notes for the editor and author (see Section 2).
• Cover illustrations. For articles without a cover, AI can create an illustration based on the article. Authors are free to use it, replace it with their own image, or have no cover at all.
• Event recording transcripts. AI listens to event recordings uploaded by organisers and writes them out as text. Transcripts are checked by our team before they appear on the event page.
• Translation. AI can produce draft translations of Knowledge Base articles and transcripts. These drafts are reviewed before publication.
• Search by meaning. AI reads Knowledge Base articles and posts so that search and related-article lists can find writing with a matching meaning, not just matching words.

Article review and cover suggestions run automatically when an article is submitted. Transcripts and translations can only be started by authorised team members, and their number is limited each day.

---

4. What AI reads, and what it does not

AI only reads:

• the title, summary, and content of submitted or published articles;
• images within articles and their alternative text;
• Knowledge Base articles; and
• event recordings uploaded by organisers, including the speakers' voices.

AI never reads:

• email addresses, phone numbers, passwords, or other account data;
• private messages, support tickets, or WhatsApp conversations; or
• posts that are still drafts and have not been submitted for review.

We do not use members' content to train AI models. Details about data processing are in our Privacy Policy.

---

5. How we label AI output

All AI output shown to the public carries a label that screen readers also read out:

• AI-made article covers are marked “AI-generated illustration”;
• recording transcripts include a note that they were made with AI and have been checked; and
• Knowledge Base articles translated by AI include a note that they were translated with the help of AI.

Arunika's notes during review are not shown to the public, but always appear under Arunika's name.

---

6. Limits of AI

AI can make mistakes. Transcripts may mishear names or terms, translations may sound stiff or miss the meaning, and Arunika's suggestions do not always match what the author intended. That is why every AI output is reviewed by a person before it is shown, and why we label it.

---

7. Feedback

If you find a mistake in a transcript, translation, or illustration, or have questions or concerns about how InfiArtt uses AI, please contact [email protected]. We will update this page whenever the way we use AI changes.