Public Accountability & Trust

Transparency & Moderation

A clear, candid account of how we protect users, screen content, retain data, and handle appeals.

1. Our Moderation Philosophy

Miloo was created on a foundational commitment: spontaneous human connection must be safe, civil, and private. Anonymity should empower candid dialogue, not shelter exploitation or abuse.

To uphold this balance, our moderation architecture combines privacy-preserving client-side analysis, edge-level server screening, and user-initiated reports. We never sell user data, record calls, or maintain persistent chat histories. Review our Safety Center and read our Community Guidelines for everyday user guidelines.

2. How Content Screening Works

We deploy multi-layered automated safeguards to intercept harmful content before it disrupts another user's experience:

  • Server-Side Text Screening: Every text message is processed through an automated content safety check prior to delivery. When configured with an API key, we utilize OpenAI's Moderation API; otherwise, an open-source local Detoxify model or internal rule heuristics evaluate the text. Messages containing hate speech, sexual solicitation, severe harassment, or phishing links are immediately blocked. Senders receive a strike warning; repeated violations result in immediate session detachment and temporary bans. Miloo does not use AI bots or automated partners. Every chat is with a real person.
  • On-Device Video Analysis: Video streams are sampled client-side using TensorFlow.js and self-hosted NSFWJS model weights running locally in your browser (one frame every 3 seconds). If explicit nudity or sexual misconduct is detected, the remote video is immediately blurred with a safety alert, the connection is terminated, and the incident is flagged. Frames never leave the user's browser unless the user chooses to submit a formal report.
  • Human-in-the-Loop Review: Reports submitted by users are queued in an internal moderation portal reviewed by our Trust & Safety staff to resolve contested cases and detect systemic spam networks.

3. What Is Strictly Blocked

Miloo enforces a zero-tolerance policy against the following behaviors:

  1. Child Sexual Abuse Material (CSAM): Any depiction, solicitation, or discussion involving minors is immediately blocked, reported to NCMEC (National Center for Missing & Exploited Children) and law enforcement, and results in a permanent network-wide ban.
  2. Non-Consensual Nudity & Sexual Acts: Genital exposure, masturbation, or sexually explicit behavior without mutual adult consent.
  3. Harassment & Hate Speech: Targeted racial, religious, sexual orientation, or gender-based slurs, threats of violence, and cyberbullying. Read our guide on what to do if someone harasses you online.
  4. Scams & Financial Fraud: Phishing URLs, fake investment schemes, extortion, and impersonation bots. Learn more in our guide on spotting chat scams.
  5. Doxxing: Revealing private legal names, phone numbers, home addresses, or social accounts of third parties.

4. Data Retention & Logging Policies

We believe that the safest user data is data that is never stored. Here is exactly what we keep and for how long:

Data CategoryRecipient / Third PartyPurposeRetention Period
Video & Audio StreamsMatched Peer (WebRTC P2P)Live real-time calling0 seconds (never recorded or mirrored)
Text Chat MessagesOpenAI (if key configured) or local Detoxify / Regex engine; matched peerAutomated safety check & message deliveryCleared on skip/disconnect (in-transit check only)
Bot Verification TelemetryCloudflare Turnstile (challenges.cloudflare.com)Bot mitigation & anti-spam pre-screeningEphemeral (tokens verified server-side and discarded)
Cookieless AnalyticsUmami (analytics.umami.is or self-hosted)Aggregate pageviews & Web Vitals on public pages onlyAggregate metrics; no cookies or chat content
Google Analytics 4 (GA4)Google LLC (googletagmanager.com)Aggregate traffic analysis on public informational pages only (Consent Mode v2 default denied; requires opt-in; strictly excluded from /chat, /call, /admin, and /transparency)_ga (2 years), _ga_<container-id> (2 years); 0 IP storage
Server-Side Matching CountersMiloo internal server (no third party)Aggregated matchmaking performance (% matched ≤15s/30s/60s, median wait, wait timeouts, chat length buckets, ended_by reasons, peak concurrent users)Statistical aggregates only; zero IPs, zero user IDs, zero chat content
Hashed IP & Device TokensMiloo internal storage (PostgreSQL / JSON)Enforcing temporary suspensions and permanent bansDuration of ban (10 min, 24 h, 7 days, or permanent)
User Abuse ReportsMiloo moderation queueInvestigation of reported harassment or illegal acts30 days (purged automatically)
Moderation Incident LogServer filesystem (incidents.log)Auditing fail-safe triggers and moderation outagesServer diagnostic cycle; rotated periodically

5. Escalating Bans & Enforcement

To deter repeat offenses while accounting for shared networks (such as university dorms or mobile carrier NAT gateways), Miloo applies progressive enforcement:

  • First Strike: Real-time message block and warning modal.
  • Second Strike: 10-minute temporary suspension.
  • Third Strike: 24-hour network ban.
  • Severe or Repeated Abuse: 7-day to permanent blacklist across both IP hash and device tokens.

6. How to Appeal a Suspension or Ban

Because we use automated heuristics, false positives can occasionally occur—especially on communal Wi-Fi networks where a previous user may have caused a shared IP ban.

If you believe your connection was restricted in error:

  1. Email our appeals desk at support@miloo.chat with the subject line Ban Appeal.
  2. Include your approximate timestamp, connection mode (Text or Video), and general ISP/region.
  3. Our human moderation staff reviews appeals within 24 to 48 hours and will clear mistakenly flagged identifiers.

For further legal and privacy policies, review our Privacy Policy and Terms of Service.

← Back to HomepageLearn about our team →