Model Explainer

Who can access your uploaded model. Within this app, only you: every model is tied to the account that uploaded it, and no other signed-in user can open it or chat about it, even with its session ID. Stored copies are encrypted at rest, so the storage provider — and anyone who obtained the storage credentials alone — cannot read them. Two parties still can: whoever administers this server, because the decryption key lives on it; and Anthropic, because the extracted formulas, code and constants are sent to the Claude API to generate the summary and answer your questions. Traffic is encrypted in transit and nothing is exposed to the public internet. Nothing is deleted automatically, but you can delete any of your models at any time, which removes it from storage immediately.

Plain-English rundown, for a private model you'd rather not share widely. The rating is how worried to be, not how safe: 1/10 means low risk / barely worth worrying about, 10/10 means serious. Higher number = bigger problem.

• A bot or scraper grabs your model off the web — 1/10. The page only shows a generic upload box. Your file and its summary are tied to a random session link that isn't published or listed anywhere, so an automated crawler hitting the site sees nothing of yours. The live site also sits behind a login.
• Your uploaded file runs malicious code on the server — 1/10. It can't. Excel formulas, VBA, Python and R are all read as text only; the app never opens, imports, sources or executes them.
• Someone intercepts the upload in transit — 1/10. Traffic goes over HTTPS, so it's encrypted between your browser and the server.
• Another visitor stumbles onto your session — 1/10. Every model belongs to the account that uploaded it. Another signed-in user who somehow obtained your session ID still gets "not found" — the check is on ownership, not on knowing the ID. Someone with access to your own account, or to a browser you left signed in, is a different matter.
• Your model sits on the server after you're done — 2/10. The parsed model and its summary are stored (in cloud object storage) so the chat survives restarts, encrypted at rest so the bucket alone gives up nothing. You can delete any of your models from the "Your models" list, which removes it from storage immediately. What remains: nothing expires on its own, and the decryption key sits on the same server as the data, so anyone with access to that server can still read what you haven't deleted.
• Your model's contents are sent to a third party — 4/10. To write the summary and answer questions, the extracted formulas / code and constants are sent to Anthropic's API. That's a company outside this site, subject to their retention policy. Don't upload anything you're contractually barred from sharing with a sub-processor.
• Secrets hidden inside the model leak — 5/10. If your file has a hard-coded password, API key, database connection string or personal data, the analyzer will extract it, send it to the Anthropic API, and store it with the session. Strip credentials and sensitive records out before uploading.

Bottom line: fine for ordinary business models; not for anything containing secrets, regulated personal data, or material you're not allowed to share with third-party services.