WorldAgents processes microphone audio, camera input, prompts, files, and related spatial or device context to provide the features you request, including when that input comes from a paired wearable device you choose to use. This content may be sent to AI, recognition, cloud, and device-service providers that help us operate those features. Requests to AI model providers are routed through our servers.
We do not record or store your audio or video. Microphone audio and camera input are processed live: they are sent to the AI, recognition, and cloud providers that perform the feature you are using, and we do not keep them once they have been processed.
We keep a text transcript and a technical session timeline so authorized users can review the session and we can operate and troubleshoot the service. These can include descriptions of what the camera showed and the objects that were recognized, but not the audio or video itself.
A still image is stored only when you upload one, or when you use an agent feature that has to keep a picture to work, such as remembering a view or creating an image from what your camera shows. It is kept as a file of that session.
Access to session transcripts, timelines, and files is limited to authorized organization users and to personnel and service providers who need it to operate, secure, support, or troubleshoot WorldAgents.
If your image, voice, or location information is captured through a WorldAgents deployment operated by one of our customers, that customer ordinarily determines why the information is collected and is responsible for providing any required notice and obtaining any required consent. Contact the organization operating the deployment first; you may also contact us if you cannot identify it.
Model providers may separately retain content for features such as session resumption, safety and abuse monitoring, or legal obligations, subject to the applicable provider terms and configuration. Retention by these providers may differ from our own retention periods.
We process audio, images, and session content to provide and secure the requested features, and we use the session content we keep to evaluate, troubleshoot, and improve them, including review by authorized personnel where needed for those purposes. Our AI service providers process content under their applicable terms and configuration. Use of non-public customer content to train models for general use requires authorization through a separate agreement, an enabled product setting, or written permission, with any additional notice or consent required by law.