Privacy Policy

01Scope and who we are

This policy covers the KNIGHT and AEGIS Discord applications (the "bots"), ALPHA Core, the shared layer that coordinates them, and this website. They are developed and operated by Saikat Mahara, an individual developer, under the name SAIKATxALPHA / TEAM SAIKATxALPHA ("we", "us", "our"). Saikat Mahara is responsible for the information the bots process, as described in this policy.

Discord itself collects and processes your information under Discord's Privacy Policy. This policy covers only what our bots do with the information Discord makes available to them.

02Summary

03Information we process

Which of the following apply depends on which bot is in a server and which features the server's administrators have turned on.

3.1 Discord identifiers and profile information

• Identifiers: Discord user IDs, server (guild) IDs, channel IDs, role IDs, message IDs and scheduled-event IDs. These link records to the right server and member.
• Profile and membership details provided by Discord: username, display name, avatar, roles, server join date and account creation date. They are used when needed, for example to draw a welcome card with your name and avatar, check role hierarchy, or spot new accounts during a raid. They are not copied into the database as a profile.

3.2 Message content

Both bots use Discord's message content access to read messages in channels they can see.

• KNIGHT checks messages for profanity, spam, repeated content, mass mentions and invite links.
• AEGIS reads messages to award activity points ("XP" and "aura"), detect duplicate messages, and recognise typed ALPHA commands.

This happens in memory. The bots do not keep a copy of server conversations. Limited content is stored in these cases:

• an excerpt of a message that triggers a moderation case (up to about 300 characters), kept as evidence in the server's private case log;
• short one-way fingerprints (hashes) of a member's last few messages, used to detect duplicates for activity rewards. The original text cannot be read back from them;
• parameters of ALPHA commands and staff actions, such as a reason given for a timeout, recorded in the audit log;
• research queries and the resulting digests, when someone requests research.

The bots do not read your direct messages with other people. KNIGHT may send you a direct message to tell you about a warning.

3.3 Moderation and security records

• Moderation cases: server ID, the affected user's ID, the moderator's ID (or the bot's, for automatic actions), the action, reason, rule, points, duration, evidence excerpt, whether it was simulated, the result, and the time.
• Voice review items (only when voice moderation is on): server, user and channel IDs, the transcript text, a confidence score, the matched term, review status, and the reviewer's ID.
• Audit log: privileged requests and decisions, including denied requests and approval outcomes. Each entry records the actor's ID, the action, its target, its parameters, the outcome and the time.
• Security events: raid and new-account alerts are posted to the server's security-log channel. Join and leave events are recorded with user ID and time.

3.4 Server configuration

Settings chosen by each server's administrators: log channels, moderation thresholds, enabled features, welcome and report options, trusted roles, shop items, scheduled jobs and reminders, events, and music preferences.

3.5 Activity, progression and events

• Activity counts: messages per member per day, voice minutes per member per day, and messages per channel per day. These are counts only, never message text.
• Progression: XP, aura balance, level, streaks, quest progress, inventory, purchases, season results, and an append-only history of currency and XP changes.
• Events: event details and the IDs of members who join them.

3.6 Music requests

Search terms and links you give to music commands, the server's queue and now-playing track, and the voice and text channels the player is using. The queue is saved so it can resume after a disconnect.

3.7 Research and optional AI

Research queries are sent to public news and reference sources (listed in section 6) to fetch results. Where the operator has enabled hosted AI:

• an ALPHA instruction that the built-in parser cannot understand is sent to the AI provider, together with the list of permitted actions;
• a research query and the retrieved public items may be sent to the AI provider to produce a short, labelled synthesis.

3.8 Operational logs, usage counters and backups

• Logs: technical logs record events and errors. They can include Discord IDs and error details. Bot tokens and API keys are automatically redacted.
• Usage counters: daily counts of speech-to-text and AI requests, overall and per server, used to enforce usage caps.
• Short-lived context: so ALPHA understands phrases like "that channel" or "that member", it remembers the last channel, member or role you referred to, for about 30 minutes.
• Backups: periodic copies of the database (see section 7).

04Voice capture and cloud transcription

4.1 ALPHA voice assistant (AEGIS)

• Off by default. It starts only when an authorised user runs the listen command in a voice channel. AEGIS then posts a listening notice, and its status shows that it is listening.
• Who is captured: only users the server has authorised to use ALPHA (its owner and trusted users or roles). Discord identifies the speaker of each audio stream, and AEGIS does not subscribe to anyone else's audio.
• What is sent: each utterance from an authorised speaker, up to about 15 seconds, is converted in memory to an audio clip and sent to the speech-to-text provider. This includes speech that does not begin with "ALPHA", because the wake word is detected from the transcript.
• Recognition hints: if administrators add custom vocabulary words (for example, names of games or artists) or set a language, these are sent with the audio to help recognition.
• What is kept: raw audio is held in memory only while being processed and is not written to disk by the bots. By default, each recognised command is posted as text in the server's ALPHA channel, with the speaker mentioned; administrators can turn this off. Commands are also recorded in the audit log. Speech without the wake word is discarded after transcription.
• Stopping: say "ALPHA, stop listening", run the stop command, or have the bot leave the voice channel.

4.2 Voice moderation (KNIGHT)

• Off by default. A server administrator must turn it on and start it in a voice channel.
• Who is captured: members speaking in that channel, except bots, the server owner, and members with roles the server has exempted. This includes ordinary members.
• What is sent: each speaker's utterances are sent to the speech-to-text provider, and the transcript is checked against the moderation word list.
• What is kept: uncertain matches are stored as voice review items, with the transcript, for staff review. Confident matches create a moderation case containing a transcript excerpt of up to about 200 characters. Raw audio is not written to disk by the bots.

4.3 Notice and consent

Server administrators and authorised users who turn on a voice feature must tell the people in the voice channel beforehand, and get any consent that applicable law requires (see our Terms of Service). If you don't want your voice processed, leave the voice channel, or ask the server's staff to stop the feature or exempt you.

4.4 Provider handling

Audio sent for transcription is handled under Groq's own terms and privacy commitments. Groq's documentation says it does not retain inference request data by default. See Groq's data policy, and check it for current details, because third-party policies can change. Groq is the only cloud speech-to-text provider the bots support. If no provider is configured, voice features refuse to start and no audio is sent anywhere.

05Why we process information

• To provide the features that server administrators turn on: moderation, security alerts, music, server management, welcome cards, progression, events, research, reports and the ALPHA assistant.
• To keep communities safe: detecting spam, raids and rule-breaking content, and keeping case history so staff can make and review decisions.
• To protect the services: authorisation checks, confirmations, audit records, usage caps, abuse prevention and troubleshooting.
• To meet legal obligations and respond to valid legal requests where required.

We do not use information from the bots for advertising, profiling for marketing, or selling to anyone. We do not use your messages or audio to train AI models.

06Third-party providers

The bots use the providers below. Some are optional, and are used only if the operator enables them or an administrator uses the related feature. Each provider handles data under its own terms and privacy policy.

Requests to music and research providers come from the bot's host, not from your device. They do not include your Discord ID. Hosted AI, Jamendo, internet radio, Google News and extra RSS feeds are optional. Which of them are active depends on the operator's configuration, and a provider that is not enabled receives nothing.

07Retention and deletion

This section describes how the software currently behaves. Most stored records are not deleted automatically after a fixed period, and removing a bot from a server does not automatically erase that server's stored data.

We have not set fixed retention periods for these records. If we introduce them, we will update this policy. You can ask us to delete information linked to your Discord user ID, and a server's administrators can ask us to delete that server's data, at any time (see section 9).

08Security

We take reasonable steps to protect the information the bots handle, including:

• keeping bot tokens and API keys in private configuration that is not published, and redacting them from logs;
• storing data in a local database on the system that runs the bots, not in a public or shared database service;
• using parameterised database queries, and limiting ALPHA to a fixed set of validated actions with no shell, code-execution, file-system or raw-database access;
• checking permissions, requiring confirmation for higher-risk actions, and recording privileged actions in an audit log;
• sending data to external providers over encrypted (HTTPS) connections.

No system is perfectly secure. We cannot guarantee that information will never be accessed, disclosed, changed or lost because of a security incident. If you find a security problem, please report it to us (see section 14).

09Your choices and rights

• Server administrators control which features run in their server. They can turn off voice features, change or reset settings, and remove the bots.
• Members can avoid voice processing by leaving a voice channel where it is on, and can ask server staff about how the bots are used in that server.
• Access, correction and deletion: you can ask us for a copy of the information we hold that is linked to your Discord user ID, or ask us to correct or delete it. To protect other users, we may need to confirm that the Discord account is yours. We may keep some records where we have a legitimate need, such as moderation records that help prevent abuse, or where the law requires it. If we do, we will tell you.
• Depending on where you live, you may have other rights under data protection law, including the right to complain to a data protection authority.

To make a request, use the contact details in section 14. Please include your Discord user ID and the relevant server ID, but never your password or account token.

10Children

The bots are available only through Discord, which requires users to meet a minimum age set for their country. The services are not directed at children below that age, and we do not knowingly collect information from them. If you believe a child below Discord's minimum age has information held by the bots, contact us and we will delete records linked to that account.

11International processing

Discord, Groq and the other providers may process information in countries other than yours, including the United States. Data-protection rules there may differ from those where you live.

12This website

This website is a set of static pages. It uses no cookies, analytics, tracking pixels, advertising or scripts, and it loads nothing from third-party servers. As the hosting provider, Cloudflare processes standard technical data, such as IP addresses, to deliver the pages and protect against abuse. It does this under Cloudflare's Privacy Policy.

13Changes to this policy

We will update this policy when the bots' data practices change, for example when a feature or provider is added or removed. We will post the updated version on this page and change the "Last updated" date. If a change is significant, we will try to give reasonable notice in advance.

14Contact

For privacy questions or requests, contact Saikat Mahara (SAIKATxALPHA):

• Email: saikatmahara7895@gmail.com