How we keep this actually positive

"Anonymous positive social app" sounds nice until you remember what happened to Sarahah, NGL, Yik Yak, and every other anonymous platform people have tried. The pattern is consistent: launch, viral moment, harassment problems, decline. We thought hard about this before launching GlazeTrain.

This page is the complete rundown of how GlazeTrain stays safe — the technical systems, the human moderation, what gets blocked at the door, and what you (the user) can do if something gets through.

The positivity check

Every single Glaze passes through an AI content check before delivery. The check is doing three jobs:

1. Filter the obvious stuff.

Direct insults, slurs, sexual content, threats, harassment — the AI catches the vast majority. Rejected messages get a "Hating detected 🚨" response back to the sender, with a brief reason.

2. Catch the sneaky stuff.

This is where AI moderation outperforms keyword filters. Sarcasm ("you're really brave for wearing that"), backhanded compliments ("you're much smarter than you look"), pretend-praise ("good for you for trying"), faux-supportive content with negative subtext — these all get flagged.

3. Reject "not a compliment" content.

Messages that aren't insults but aren't compliments either — random text, questions, complaints, attempts to use the platform like a DM service — get rejected with "Not a compliment 🤷" so the platform stays on-topic.

What gets rejected

Examples of Glazes that fail the positivity check:

The two-layer approach

AI moderation is the first layer. It's fast, runs on every Glaze, and catches the bulk. But AI moderation has known failure modes — context confusion, edge cases, novel phrasings. So we built a second layer: human moderators who review reports.

When you tap "Report" on a Glaze, the report goes into a human review queue. Goal: every report is reviewed within 24 hours by a person, not an algorithm. If the human moderator agrees the Glaze shouldn't have passed, the sender's account gets sanctioned.

Sanctions, in escalating order:

• Warning. First-time or borderline cases. Account is notified that the content was found to violate guidelines.
• Anonymous mode revocation. If anonymous abuse is the pattern, the account loses the ability to send anonymous Glazes.
• Temporary suspension. Account is locked from sending Glazes for 24 hours, 7 days, or 30 days.
• Permanent ban. Account is closed and can't be re-created from the same device or phone number.

Block: your individual control

Independent of platform-wide moderation, every user can block individual senders:

• Tap (···) on any Glaze → Block sender.
• This works on anonymous senders too — the platform still knows the account behind the mask.
• Blocked senders can't send you Glazes (anonymous or signed), challenge you to battles, or join Glaze Trains that include you.
• Blocks are silent — the blocked user gets no notification.
• Reverse anytime via Settings → Blocked Users.

What about groups being targeted?

One thing we learned from anonymous-platform history: even compliments can be used as harassment when they're targeted by protected characteristic. "You're so articulate for [demographic]" sounds positive, isn't, and AI moderation has to be specifically tuned for that.

Our AI check is configured to flag any Glaze that references a protected characteristic — race, religion, gender, sexual orientation, disability, national origin, age — even if the surface content seems positive. Most of those get rejected outright; the close calls get queued for human review before delivery. This is one of the few places where we err strict.

What about minors?

GlazeTrain is rated 13+ on the App Store and Google Play. Accounts created with under-13 birth dates are rejected at signup. We don't accept any content that targets, references, or sexualizes minors — there's a hard line and it doesn't move.

If you encounter content that violates this, report it immediately. CSAM and grooming-related reports are escalated to a senior moderator within 1 hour and reported to the relevant authorities per our child safety policy.

The limits of moderation

We won't pretend moderation is perfect. No platform's is. Things that even our two-layer system might miss:

• Inside jokes that read mean. A Glaze that's between friends might look bad to a moderator unfamiliar with the relationship.
• Compliments with hidden context. "Your hair looks great today" can be a passive-aggressive jab in the right friendship dynamic — but the AI sees only the text.
• Coordinated abuse. Individual Glazes that pass but, in aggregate, are clearly someone being weird about a specific person — these need pattern detection that's harder to get perfect.

For all of these, the report button is the right move. Reports trigger a human review with the full sender history, which is where these contextual cases get resolved.

If you're being harassed

Worst case, despite all of the above, something gets through that shouldn't have. Here's the playbook:

1. Block the sender immediately. Stops further contact from that account.
2. Report the Glaze with as much context as you can include. Reports are reviewed within 24 hours.
3. Turn off anonymous Glazes if the harassment was anonymous. Settings → Privacy.
4. Email support at support@glazetrain.com for serious or repeated harassment that crosses platform lines. We respond within 24 hours.
5. If the content is threatening or illegal, contact local law enforcement. We cooperate with valid law enforcement requests under our Privacy Policy.