Pre-release Tensor Cortex Studio builds keep the canonical project, imported datasets, prompts, targets, predictions, model files, adapters, checkpoints, experiment history, and exported artifacts on the user's Mac by default. The application does not include product telemetry, automatic crash reporting, accounts, or remote project storage. A folder you place in iCloud Drive, another sync service, a network volume, or a backup remains subject to your operating-system and provider settings; Studio does not control those independent transfers.
When a user chooses to obtain a supported model, Studio connects to the model repository or provider identified in the confirmation screen and requests the selected files and revision. Current pre-release builds use Hugging Face. Tensor Cortex does not send the user's datasets, prompts, outputs, model weights, project files, or credentials with that request. The selected provider can nevertheless receive ordinary connection data, including IP address, device or software information, request time, and the requested repository files or revision, under its own privacy policy and terms.
Studio uses Sparkle to request signed update metadata from tensorcortex.com and immutable update files from downloads.tensorcortex.com. A user may check manually. Scheduled checks run no more often than every 24 hours only after the user opts in and may be disabled at any time. Downloading and installing an update remains a separate choice; silent installation and Sparkle system profiling are disabled.
Studio includes an offline model catalog and can save model shortcuts only on your Mac. Opening or searching that list does not contact us. Only choosing Update Model List requests a small signed JSON catalog from models.tensorcortex.com, served by our hosting provider. Its URL identifies your Studio version and build; the provider also receives ordinary connection information such as IP address, time, protocol and a generic user agent. The request contains no account or installation identifier, cookies, credentials, hardware profile, project data or selected-model information. There is no automatic or periodic catalog check, and we do not use catalog requests to track application usage. Updating the catalog does not download model weights or change project selections. Removing a local shortcut does not delete the model files. Catalog requests are not anonymous or invisible to the delivery infrastructure; hosting and security metadata remains subject to the provider arrangements described in this Policy.
Update requests expose ordinary connection information to our hosting and update-delivery infrastructure, such as IP address, request time, requested URL, protocol, and user-agent or software information. Studio does not attach project names or paths, datasets, prompts, targets, outputs, predictions, model files, adapters, checkpoints, credentials, or a Tensor Cortex account identifier. Publicly distributed builds will use Apple Developer ID and notarization for the application, while Sparkle Ed25519 signatures separately protect update metadata and archives.
Future local features, including additional model providers and local response distillation, do not by themselves authorize Tensor Cortex to receive project content. If a feature would send teacher prompts or responses, datasets, models, credentials, project content, or other personal information to a remote model API or service, Studio must identify that recipient and the data sent before the user enables or submits the operation.
If a future feature would transfer project content or introduce a new recipient, an in-product notice and any required supplemental terms will identify the operator and processors, data categories, purposes, legal bases, regions and transfer safeguards, retention and deletion rules, security boundaries, and available controls before use. This Policy provides information; it is not a request for blanket consent. Reading this Policy or accepting the software license, an update check, or a model download does not authorize another transfer of local project content.