Chamet Strengthens Multimodal AI Moderation for Live Video Chat as Singapore’s Online Safety Commission Takes Effect

Chamet, a global live streaming and video chat platform, is enhancing its multimodal AI moderation system to identify and address harmful content in real time. Operating primarily in Southeast Asia, where its user base is strongest, Chamet has incorporated multimodal detection of suspected minors, ensuring that only adults aged 18 and over can access the platform.

The system evaluates profile photos, live-stream frames, video, voice, text, self-reported age, and account behavior, assigning a risk level to each case. Human reviewers then assess these cases before final decisions are made, with accounts found to have falsely claimed their age facing permanent suspension.

Chamet employs multiple moderation models and third-party services, each with unique labels and confidence scales. These outputs are standardized onto a single risk scale, allowing the platform to weigh them based on factors such as content type, where the content appears, and each model's historical performance. This multi-model fusion process enables a consistent action to be taken against the highest-risk cases.

The systems are also paired with real-person face verification, measures to prevent unauthorized screen recording, in-app reporting and blocking, and an appeals process.

Additionally, Chamet offers free real-time AI translation in over 50 languages, converting speech into on-screen subtitles during video calls. Speaking on the platform's commitment to safety, Chamet's Chief Marketing Officer, David Qi, emphasized that "in a live video call, harm can happen in seconds, so safety has to work just as fast."

By integrating numerous models and signals, Chamet's review teams can detect risk more swiftly and ensure more consistent decisions, ultimately balancing the act of connecting people with safeguarding their well-being.