Security and privacy
The most dependable way to protect a conversation is to keep it where it happens. bitHuman avatars can render on the device, on your own servers or completely offline, so you decide what — if anything — reaches us. This page sets out what we receive in each mode.
When the avatar renders on your hardware and the conversation runs on your own services, bitHuman receives usage metering only — never audio, video or conversation text.
Self-hosted and on-device sessions check your credential when they start and report usage while they run, so your credits can be metered.
Usage reports contain no audio, video, images or conversation text. Self-hosted and on-device sessions store no transcript at bitHuman.
An avatar animates from audio. Where that audio comes from — and where speech recognition, the language model and the voice run — is your choice: the CLI's local conversation brain on macOS and Linux, which keeps all three on the machine; your own speech services and any OpenAI-compatible model endpoint, including one inside your network; or bitHuman's cloud voice service.
Only bitHuman's cloud voice service sends conversation audio and text to bitHuman and its service providers. With the web embed, the conversation runs on bitHuman's servers, even when the avatar renders in the tab.
Creating an avatar from a portrait happens in the bitHuman cloud; the portrait is uploaded for that step, in every mode. The finished avatar model then downloads to your devices and runs there.
Connections to bitHuman use HTTPS (TLS). Real-time media travels over WebRTC, which encrypts audio and video in transit (DTLS-SRTP). Provider keys you connect are encrypted at rest.
Organization roles (owner, admin, member); an audit-log API; API-secret rotation that revokes the old secret immediately; and scoped runtime and embed tokens, so browsers never hold your secret.
The Apple and Android SDKs hold an API secret on the device: give each app or device fleet its own secret that you can rotate or revoke.
Sessions that render on your hardware store no transcript at bitHuman. Conversations in the bitHuman cloud are stored with your agent so you can review them.
Deleting an agent deletes its records, including transcripts, and its model files.
From our privacy policy:
“We do not use your content to train our AI models unless you explicitly opt in.”
“We do not sell your personal information to third parties.”
HIPAA, GDPR and CCPA obligations follow the data, so we start with architecture.
In on-device, self-hosted and offline deployments, patient and customer conversations stay in systems you already govern, and your existing programs apply.
Healthcare and financial-services deployments are set up under an enterprise agreement and review. We are glad to complete security questionnaires and walk your team through each data flow.
When the avatar renders on your hardware, its audio and video stay there. The conversation runs where you choose: the CLI's local conversation brain, your own speech and language services, or bitHuman's cloud voice service. With the first two, bitHuman receives usage metering only — never audio, video or conversation text. In the bitHuman cloud, and with the web embed, audio and conversation are processed by bitHuman and its service providers to run the session.
HIPAA compliance is a property of your program and deployment, not of a single software component. bitHuman can be deployed so that no protected health information reaches bitHuman — on the device, on your own servers with speech and language services you control, or fully offline — and sessions that render on your hardware store no transcript with us. Healthcare deployments are set up under an enterprise agreement and review; contact us to walk through your data flows.
With on-device, self-hosted or offline deployment, your customers' conversations stay in the systems and regions you choose, and bitHuman receives usage metering only. Creating an avatar from a portrait happens in the bitHuman cloud. We do not sell your personal information, and we do not use your content to train our AI models unless you explicitly opt in.
No. We do not use your content to train our AI models unless you explicitly opt in.
Avatars in the bitHuman cloud render in the US. If the avatar must render somewhere else, render it on the device or on your own servers.
