Does Filmsquare train AI on my work?
No. The terms say: "We do not train AI models on your customer content, and we do not let others do so on our behalf." The data processing agreement adds that we do not sell customer content and do not process it for our own purposes.
What we do use it for #
You give us a licence to host, store, copy, process and transmit your content, to send it to AI providers, to make renditions of it (such as thumbnails, proxies and previews), to show it to the members of your workspace and to guests through share links, to use it for safety, moderation and abuse prevention, and to meet our legal duties. We use that licence only for these purposes and to provide, secure and support the service for you. A voice clone or other custom model you ask the service to make from your content is made for you, and is not training.
What is sent to AI providers #
When someone generates, we send the prompt, settings and any reference material (such as images, audio or text from the workspace) to the AI provider that runs the model, and receive the result. A provider receives this only when a generation runs on one of its models. The privacy policy names them: BytePlus, Kling, Google (Gemini API), fal.ai, Topaz Labs, Runware, ElevenLabs, Higgsfield AI, a GPU hosting provider we rent for our own model workflows, and Anthropic only when someone asks for an AI breakdown of a script and that feature is switched on.
AI providers process the content under their own terms. Where a provider has told us it does not use API content for training or to improve its products, the privacy policy says so, as it does for Google's paid service and Topaz Labs. For the others, their API terms decide, and we make no promise beyond those terms. The privacy policy also says how long each provider keeps what it receives, and you can ask us for the terms of a specific provider.