Community Policy

The Community Guidelines define conduct on Alpha. This policy defines process: who reviews content, how decisions are made, how the platform's four moderation pillars work, and what rights you have at every step.

1. Purpose

Alpha is a network where founders and builders share unfinished work and honest numbers. That only survives if people trust the environment. Trust requires moderation that is fast, consistent, explainable, and reversible when wrong. This page is the public record of how we pursue that.

2. The four pillars

Following modern platform practice, our approach rests on four actions:

1. Remove violative content — anything crossing the Guidelines' "not allowed" lines.
2. Restrict borderline behavior — posting limits, link limits, and enhanced review rather than removal where context matters.
3. Review — automation handles scale; humans handle context, reports, and every appeal.
4. Empower members with real tools: blocking, muting, reporting, private accounts, approval-based communities.

3. Definitions

• Content — anything a member creates: threads, replies, images, profiles, community metadata.
• Violation — content or conduct breaching the Guidelines.
• Strike — a formal recorded mark against an account for a confirmed violation.
• Screening — automated pre-publication review of post text.
• Review — human evaluation of reports, flags, and appeals.
• Restriction — temporary limits on posting, linking, reacting, or messaging.
• Removal — deletion of specific violating content while the account remains.
• Ban & deletion — permanent loss of access plus erasure of the account's data.

4. Who moderates

4.1 Automated screening (first layer)

Before publication, thread and reply text passes through two machine checks: a deterministic screen for profanity, slurs, sexual terms, self-harm phrasing, threats, scam patterns, exposed personal data (emails, phone numbers, card numbers), link floods, and manipulation patterns; then AI classification against safety categories — violence, hate, harassment, sexual content, self-harm, spam, privacy violations.

Screening completes in seconds. Passing text publishes normally; flagged text is blocked, and the author is told which category triggered the block.

4.2 Human review (second layer)

Context-sensitive cases—such as reports about images or messages, borderline calls, and appeals—are intended for human review when available. Review applies these written policies and should, where reasonably possible, be handled by someone other than the original decision-maker.

4.3 Community owners (third layer)

Community creators run their spaces within platform rules — extra rules are welcome, looser standards are not. Owners can remove off-topic posts and control membership; they cannot grant exceptions to the Guidelines, and communities organized around violating them are removed entirely.

5. What gets screened — and what doesn't

• Machine-screened: text of threads and replies before publication.
• Not routinely pre-screened by the current text classifier: private direct messages, uploaded images, and profile fields.
• Messages may be reviewed when reported, required for support, necessary for security, or legally compelled. See the Privacy Notice.

6. Reports

• Any member can report any post, reply, profile, image, or community from the in-app menu, choosing a reason category that routes the case.
• Serious categories — threats, child safety, credible fraud, doxxing — are prioritized above all other queue items.
• Reporters aren't told outcomes on another member's account; bulk false-reporting is itself sanctionable under integrity rules.
• Reports from community owners about their own spaces carry owner context and are reviewed accordingly.

7. Deciding severity

Reviewers weigh five factors consistently across cases:

1. The rule broken — each Guideline maps to a base severity.
2. Degree — a slur aimed at a person outranks casual strong language; a scam network outranks one sloppy spam post.
3. Intent signals — bait, satire, mistake, ignorance, or malice.
4. Context — quoting to condemn differs from producing; teaching about scams differs from running them.
5. History — existing strikes raise consequences for new violations.

8. Actions available

• No action — content stands; reporter mistaken.
• Removal — specific post deleted; account continues.
• Warning + strike — removal plus recorded strike with written explanation.
• Restriction + strike — posting/link/interaction limits added.
• Permanent ban + deletion — third strike or severe single violations.
• Emergency freeze — immediate suspension pending review where ongoing harm is plausible.

9. The three-strike system

Most violations earn strikes: first removes content and warns, second adds restrictions and closer review, third bans the account permanently and deletes it from the platform — profile, posts, media, communities, connections. Full mechanics live in Strikes & Appeals.

10. Zero-tolerance categories

The following may bypass the ordinary strike ladder and result in immediate suspension, ban, preservation, or deletion:

• Child sexual abuse material or exploitation of minors in any form;
• Credible threats of violence or mass harm;
• Terrorism promotion, recruitment, or financing;
• Coordinated fraud operations targeting members;
• Doxxing campaigns;
• Ban-evasion infrastructure.

Authorities are notified where law requires or permits. Appeals remain available even here, but accounts stay suspended during appeal.

11. Ban evasion

Re-registering after ban or deletion is prohibited. Evasion accounts are removed on detection without separate process — the underlying case already received full due process. Devices and identifiers associated with evasion networks may be blocked.

12. Deletion and member data

Banned-and-deleted accounts pass through the same erasure pipeline as voluntary deletion, per the Privacy Notice. We retain minimal enforcement records — identifiers sufficient to detect re-registration — where necessary to keep banned individuals off the platform.

13. Errors and honesty

Automation will sometimes be wrong; so will humans. Every action carries an appeal right. Reversed actions restore content, lift restrictions, and erase strikes completely — as if they never happened. Our bias is explicit: we would rather wrongly allow ten borderline posts than wrongly delete one good builder's account.

14. Changes

Material changes are announced in-app before taking effect. Conduct is judged against the Guidelines + Policy pair as written at the time of the alleged violation.

15. Contact

Policy questions: safety@alpha-app.online. Enforcement specifics: Strikes & Appeals.