Trust and safety on Avatok
How we think about content policy, age gates, harassment, fraud, and takedowns — written so creators and viewers know what we will and will not allow.
Avatok is a platform where strangers meet on video to do commerce. That sentence is the whole reason trust and safety matters here in a way it doesn't on a static publishing platform: real money changes hands, real time gets booked, and the meeting is one human in front of another. We take that responsibility seriously, and this post describes how.
For the formal version of any specific policy, see the trust & safety policy. For the narrative version, keep reading.
The four pillars
Everything we do on the safety side is in service of four goals:
- Money moves to the right person. The booker pays; the creator who delivered the work receives. Fraud, scams, and unauthorized account access are the most concrete way that breaks, and they get the most attention.
- Sessions are between consenting adults. We enforce age gates on the platform and on every paid session. We do not allow content that targets, exploits, or sexualizes minors. There is no edge case here.
- Harassment doesn't pay. Creators don't have to put up with abuse to use the platform; viewers don't have to put up with predatory creator behavior. Reports are actioned by humans, and the penalties for harassment scale up quickly.
- The platform doesn't become a vector. We are unusually careful about CSAM detection, takedown requests under DMCA, harmful drug content, and the standard list of platform- abuse vectors. None of this is an acceptable use of Avatok and we will not tolerate it.
How content moderation actually works
We run a layered moderation system rather than a single moderator- per-stream model. The layers are, roughly:
- Pre-publication checks. Listings, profile photos, channel art, and uploaded videos go through automated content classification before they reach the public marketplace. The bar here is "safe-for-work, lawful, accurately described."
- Live-stream sampling. We sample short frames from live broadcasts at a configurable cadence and run them through the same classifier stack. Streams that exceed risk thresholds are surfaced to a human moderator queue in real time.
- User reporting. The fastest way for us to catch a policy violation is for someone in the room to report it. Every paid session has a one-tap report button; reports go to the trust and safety queue and are triaged inside 24 hours.
- Trust signals. Repeat reports, refund patterns, chargebacks, KYC failures — these feed a per-account trust score that controls what features you can use. Low trust shrinks the surface area until we've had a chance to look.
Age gates and verification
Every viewer who creates an account asserts they are at least 18, and we require KYC-style age verification before paid features unlock. Every creator goes through the same KYC at onboarding. The ID verification is provided by Stripe Identity — the same service we use for payouts — so a single check covers both the payment rails and the age gate.
If we determine through any signal (ID review, behavior, report) that a participant is under 18, the account is terminated, any balance is held pending review, and the case is escalated to law enforcement where applicable. There is no warning step for this.
What gets you suspended
We try to be specific about behavior that ends the account, rather than rely on vague "community standards" copy:
- Any sexual content involving minors, or that could reasonably be read as targeting minors. Zero tolerance, no appeal.
- Threats of violence, doxxing, stalking, or coordinated harassment campaigns. Permanent ban on first confirmed report.
- Selling regulated goods or controlled substances through any Avatok product surface. Listings removed, account suspended, referred to authorities where required.
- Fraud: chargeback farming, fake bookings, account takeovers, payment-rail abuse. Account suspended, funds held pending investigation, reported to Stripe.
- Repeat copyright infringement under DMCA. Three strikes and the account is closed; the strike clock can be reset under specific repeat-infringer policy criteria.
Reporting something
If you see content or behavior on Avatok that violates the policy above, the report flow is:
- On a live stream or session: the in-room report button. Goes straight to the trust queue, includes the session id and timestamp.
- On a listing or profile: the report button on the listing or profile page. Same destination, different metadata.
- By email: trust@avatok.ai. Use this for anything that needs context the in-product reports can't capture — DMCA, legal threats, off-platform safety issues.
We commit to acknowledging reports inside 24 hours. Resolution timing varies; clear policy violations are usually closed inside 72 hours, harder calls take a few days. We tell you when we've finished and what we did.
The cases that get hard
Most content moderation work is not the easy cases; it's the ones where reasonable people disagree. We'd be lying if we pretended we got every call right. The kinds of cases that come up most often:
- A creator and a paying viewer have a dispute about what was promised in the session. This is a content issue (was the work delivered?), not strictly a safety issue. We mediate; we lean toward the side that has the documented evidence.
- Sexually explicit content that is consensual and adult but that we don't allow on Avatok. We enforce. We're not the platform for that work and we're honest about why.
- Bookings that look like fraud but turn out to be a viewer experimenting with the platform. We err toward giving the user benefit of the doubt the first time and a clear warning the second.
If your case is in one of those gray areas and you think we got it wrong, write to trust@avatok.ai. A second human will look. We change decisions when we're wrong.