Moderation Transparency — How Chatzyo Actually Handles Reports

A lot of platforms publish a "transparency report" full of big numbers — total reports processed, bans issued, a false-positive rate carried to one decimal place. It looks rigorous. The problem is that for a platform our size, those numbers usually mean less than they appear to. A report count without context tells you nothing about whether the system actually works, and a false-positive rate with no audit behind it is just a number someone decided to publish.

So instead of inventing statistics we don't actually have, here's something more useful: exactly how moderation works on Chatzyo today, what's automated, what a real person actually looks at, and where the gaps still are. If a number genuinely exists for something below, we've written it. If it doesn't, we've said so instead of filling the space with something that sounds authoritative.

Why there's no report count here yet

Chatzyo was founded in 2025. We don't have a year of clean, audited moderation data sitting behind this page, and publishing a number without that audit trail would be exactly the kind of transparency theater this post is trying to avoid. When we have enough volume and history to publish numbers that actually mean something — not just a count, but context on what was reviewed, what was overturned on appeal, and how the categories break down — we'll publish them here. Until then, this page describes the system itself, which is the part we can be fully honest about right now.

What's automated

Chat messages pass through a filter before anyone sees them. It's built to catch attempts to share contact details — email addresses, phone numbers, and the deliberately disguised versions people use to get around simple filters, like writing a number out in words or substituting look-alike characters. This isn't a content-moderation AI making judgment calls about what's appropriate; it's a narrower, rules-based system aimed specifically at the kind of off-platform contact-sharing that tends to precede the harder-to-moderate problems.

We're deliberately not publishing the exact patterns it checks for. Describing a spam filter's rules in detail is mostly useful to people trying to get around it, not to people trying to understand whether the platform is safe. What we can say plainly is what it's for: reducing the number of conversations that move off-platform before anyone has a reason to trust the other person, which is one of the more common ways anonymous chat gets misused.

What a real person actually reviews

Every report submitted through the in-chat Report button goes to a human moderator, not just an automated queue that auto-resolves itself. Our stated target is a 2-hour review window, and that target governs how the small moderation team is staffed — it's a commitment we hold ourselves to, not a marketing line decoupled from how the team actually operates.

Because Chatzyo doesn't use accounts — there's nothing to log into, no profile to delete — enforcement happens at the IP and device level instead of the account level. That has a real tradeoff worth naming honestly: it's harder for a determined bad actor to permanently evade a ban by simply registering a new account, because there's no account to register. It's also true that IP-based enforcement isn't perfect — a banned IP can change on some connections — which is a limitation, not a hidden strength, and we're not going to pretend otherwise.

How enforcement actually escalates

The same enforcement table that governs our Terms of Service is the one moderators actually use — there isn't a separate internal version with different rules:

• First-time minor issues (spam, etc.) — a warning, with the session ended.
• Repeated or moderate violations — a temporary IP ban, ranging from 24 hours to 30 days depending on severity.
• Explicit content, harassment, or illegal activity — a permanent IP ban, appealable only by email.
• Automated access, scraping, or bot use — a permanent IP ban.
• Suspected underage users — an immediate permanent ban, with no appeal process, given the severity of the concern.

Every one of these is a human decision, not an automatic trigger based on a report count. A single credible report of something serious is treated as seriously as it should be — we don't wait for a pattern to form before acting on a clear violation.

What we deliberately don't do

Chatzyo doesn't record video or audio at any point — calls are peer-to-peer over WebRTC and never pass through our servers in a storable form, which is also why we can't review a flagged call after the fact the way a platform that records sessions could. That's a genuine limitation of building the product this way, and it's worth naming directly: we are trading the ability to review historical evidence for not retaining sensitive recordings in the first place. We think that's the right tradeoff for a platform like this, but it is a tradeoff, not a pure upside.

Text messages are kept only in server memory for the duration of a session and are not retained afterward, which is described in full in our Privacy Policy. We also don't use IP addresses for advertising or sell them to third parties — security logs containing IP addresses are kept for a maximum of 30 days before automatic deletion.

Where this leaves us, honestly

This isn't a finished system, and we're not claiming it is. It's a small platform with a real moderation commitment, a working enforcement policy, and an honest gap where the published statistics other platforms lead with would normally go. We'd rather describe the actual system accurately than publish a number that looks credible but isn't backed by anything you could verify.

If you've had a bad experience that the system above didn't handle well, the most useful thing you can do is tell us directly — through the in-chat Report a User flow for anything urgent, or Contact for anything else. We read those.