Privacy Policy

The short version

The current Run Any AI local labs are designed so the text or audio you choose is processed in your browser. We do not require an account or AI API key, and the application does not send your lab input or model result to a Run Any AI server.

Device checks

The model matcher checks browser-provided capabilities such as WebGPU, WebAssembly, CPU-thread support, reported memory, and estimated browser storage. These checks and the optional CPU benchmark run on your device. The site does not send the capability result or benchmark score to a Run Any AI server.

Model downloads and local storage

When you choose to run the demo, your browser downloads public model and tokenizer files from Hugging Face. That service receives ordinary network-request information, such as your IP address and browser details, under its own privacy terms. Your entered text is not included in the model-file request.

The downloaded files may be kept in browser-managed cache storage so the model can be reused. You can use the site’s removal control or clear this site’s data in your browser settings.

Preferences and selected files

If you save a model in the matcher, its identifier is stored in local browser storage on your device. Audio selected in the transcription lab is opened through a temporary browser URL, decoded in memory, and not uploaded by the application. Reloading or closing the page clears in-memory inputs and results.

Local Tone Reader extension

The developer-preview extension uses temporary active-tab access and the browser scripting interface to read only the text selection when you request that action. Selected text and classification results are processed inside the extension. The extension requests access to public Hugging Face hosts so it can download model and runtime files; those hosts receive ordinary network-request information under their own privacy terms.

Agent blueprint builder

The agent blueprint builder processes the name, objective, permissions, actions, credentials, retention choices, and generated configuration inside the browser tab. Run Any AI does not receive or store this information. Copying or downloading the generated JSON is an explicit action handled by your browser.

Compute advisor

The compute advisor processes workload type, model size, precision, concurrency, operating preferences, and storage choices inside the browser tab. The memory estimate, provider ranking, and downloadable brief are generated locally. The current release does not request provider credentials, query provider accounts, or provision infrastructure.

Hosting and security data

Like most websites, our hosting infrastructure may process basic request data such as IP address, user agent, requested page, timestamp, and security events. This data may be used to deliver the site, diagnose problems, prevent abuse, and protect the service.

Cookies, accounts, and analytics

The current release has no user accounts, advertising trackers, or application analytics. If this changes, we will update this policy and provide appropriate choices before collecting new categories of data.

Your choices and contact

You can avoid downloading a model by not starting the demo, and you can remove cached files at any time. Run Any AI does not yet have a domain email address. A verified privacy contact will be added here before the site accepts public accounts, submissions, or payments.