AI Transparency
Last updated: 29.10.2026Some features of AlphaCloud use artificial intelligence to generate or review content. This page explains which features use AI and why, in line with the transparency obligations of Regulation (EU) 2024/1689 (the EU AI Act), in particular Article 50.
How AlphaCloud Is Built
For context, even though this is not a direct AI feature within the product: the technical foundation of AlphaCloud was written by hand. The majority of the remaining code, around 90 percent, was created with AI assistance. Certain security relevant or critical areas additionally receive closer review by humans.
Overview
AlphaCloud uses a third-party large language model to power several optional features across the Discord bot, the web panel, our documentation, and our Ideas and Bug Report platforms. This page lists every feature that uses AI and the specific transparency duty of the EU AI Act that applies to it. This page does not replace our Terms of Service or Privacy Policy and is not legal advice.
AI Chat
The optional AI Chat feature lets you talk to an AI assistant in designated Discord channels, either on our main community server or on any server that has enabled the feature. Replies in these channels are generated by an AI system, not written by a human. Article 50(1) of the EU AI Act gives you the right to know when you are interacting with an AI system. If a server you use has AI Chat enabled, replies from the bot in the configured AI Chat channel are AI generated.
AI Ticket Support
When AI support is active in a support ticket, the assistant identifies itself as AI supported at the start of the conversation, and a "Disable AI" option is available at any time to hand the ticket back to a human team member. Ticket summaries created with the "/ticket summarize" command are also generated by AI. This reflects the direct interaction disclosure required by Article 50(1) of the EU AI Act.
AI Server Builder
The AI Server Builder feature, available in the web panel, generates a complete Discord server structure (roles, channels and permissions) from a text description you provide. The result is a machine generated proposal that you review and apply yourself. A "destructive" apply mode can fully replace an existing server structure and should be used with care. Presets shared through the built-in marketplace are labelled as AI generated where displayed. Article 50(2) of the EU AI Act requires outputs of a content-generating AI system to be identifiable as artificially generated. This feature is designed around exactly that: an AI-authored proposal for you to review, not an autonomous action.
"Ask Alfi" (setup assistant in the webpanel)
In the webpanel, the "Ask Alfi" assistant helps you set up your server. Its answers are generated by an AI language model (Claude by Anthropic), not written by a person. The first message of every chat and the note in the chat window's header point this out. The assistant can read the settings of the modules as well as the server's channels and roles, and suggest changes based on them: settings, message designs (embeds and Components V2), entries such as ticket categories and panels, and posting a message. It never carries anything out itself: every suggestion appears as a card showing the current and the new value or a preview and the individual steps, and only your click on "Apply" carries it out, with the same permissions and checks as the form in the panel. Steps that post a message in Discord and any links included are marked separately on the card. Answers and designs can contain mistakes, so check suggestions before you apply them. If you apply a design, the bot posts it as a message of your server. This corresponds to the obligation to disclose direct interaction under Article 50(1) of the EU AI Act. Which data is transmitted in the process is described in section 9 of our privacy policy.
Documentation: "Ask Alfi" and machine translation
On alphacloud.bot/docs, the "Ask Alfi" assistant answers questions about AlphaCloud. Its answers are generated by an AI language model (Claude by Anthropic), not written by a person. The "AI" badge in the chat window's header and the first message of every chat point this out. The assistant relies solely on our documentation. It neither sees your server nor its settings and cannot change anything. Its answers link to the relevant pages of the documentation; if an answer differs from them, the linked page applies. Answers can contain mistakes. Asking requires signing in with Discord, and a limited number of questions is possible per day. Which data is transmitted in the process is described in section 9 of our privacy policy. We write the documentation in German. All other language versions are machine translated with the help of AI (Claude by Anthropic) and are not written by a human translator. We point this out here; the individual pages carry no separate note. An automatic check ensures that every translation contains the same sections, steps, notes and links as the German page. If a translation is missing or no longer up to date, we show the English or the German version instead, with a note. The German version always prevails. This corresponds to the obligation to disclose direct interaction under Article 50(1) and the transparency principle for generated content under Article 50(2) of the EU AI Act.
Applications and competency check
In the test at alphacloud.bot/apply, for applications and for our team's competency check, an AI language model (Claude by Anthropic) plays the role of a user in short support conversations. The notice before the test and the label "AI plays the user" on every conversation point this out. The conversation partners are made up, every conversation goes differently. The AI assesses nobody and makes no decision. A dictionary on our own server counts unknown words, people from our team rate tone and knowledge, and a person always decides on applications and test results. Only the helper test is scored automatically, and without AI: it consists only of multiple choice questions with fixed points, and a person can review the result at any time. The team can let the AI suggest new knowledge questions from our documentation. Such drafts stay inactive until a person has checked and approved them. Which data is sent is described in section 19 of our privacy policy.
Automated Review and Translation
Submissions to our Ideas platform (ideas.alphacloud.bot) and our Bug Report platform (bugs.alphacloud.bot) are automatically checked by AI for quality, relevance and compliance with our content rules before publication, and matched against existing entries to avoid duplicates. If your submission is rejected or merged into an existing entry, the reason shown to you may be written by AI. Submissions are also automatically translated by AI between German and English so they can be read in both languages. The translated version is not written or reviewed by a human translator. This reflects the content-generation transparency principle of Article 50(2) of the EU AI Act.
Promotion review
When a server applies for a promotion with us, we review it once with the help of AI. For that, the last 50 messages of the channels visible to us, plus of the channel that holds our promotion, are sent to our AI provider, and only the message text, without the names or IDs of the people who wrote it. This only happens when the applicant permits it twice in two separate steps in the form and confirms that they are authorised to do so; without those confirmations no review takes place. The legal basis for this, and how the members of the reviewed server are informed about it, is set out in section 9 of our privacy policy. The AI judges three things: whether problematic content is present, what kind of server this is, and whether the promotion channel meets our requirements. Only that result is stored and, when the result is negative, the reason for it. We do not store message content. The result is a machine generated assessment and is labelled as such in the ticket. Accepting or declining is always decided by a human on our team, never by the AI alone. This meets the labelling duty for generated content under Article 50(2) of the EU AI Act.
No High-Risk or Prohibited AI Use
AlphaCloud does not use AI for biometric identification, emotion recognition, social scoring, employment decisions, creditworthiness assessment or any other purpose listed as high-risk in Annex III of the EU AI Act, and does not use any AI practice prohibited under Article 5 of the EU AI Act. This statement reflects the AI functionality present in AlphaCloud as of the date at the top of this page.