Privacy notice

Image Vault operates the Deep Scan trial. For anything in this notice, write to contact@imagevault.ai.

Nothing about you by default. Running a scan needs no account and no email address. We set no cookies on this surface.

Your IP address, held briefly in a rate-limiting counter so one address cannot flood the service. Counters expire within ten minutes and are not joined to anything else.

Visit counts. Two kinds, neither of which identifies you. Cloudflare, which hosts the site, counts page views with its cookieless Web Analytics beacon. Separately, we count how many distinct people saw the weekly chart, picked a name, viewed results, and asked for access. That count keys each visitor by a one-way hash of your IP address and browser signature mixed with a secret and the current chart week. The hash changes every week, cannot be reversed, and is never joined to anything else; the IP address itself is not stored.

Reference photos or 3D scans you choose to upload. These are optional. They must show the person being scanned, are screened by an automated check before they are accepted, and are used for one purpose only: to anchor that subject's identity during the sweep you launch. Every image upload is a photograph of a person's face, so we treat it as sensitive and never show it to any other visitor.

A note in your browser recording that you have already watched a run's replay, so a reload skips it. It stays on your device.

A results page lists content found on public platforms that the pipeline matched to the subject and assessed as likely AI-generated. Each entry carries the link to the original post, the posting account's public handle, and the detector readings behind the call. Those readings are automated assessments, not findings of fact about the account that posted the content.

Results for a subject are shared: everyone who picks the same name in the same fortnight sees the same run. Reference material is never part of what is shared.

Drafts you do not launch are hard-deleted thirty minutes after they are opened, uploads included.

Launched runs, with their results and any reference material behind them, are kept while the subject is on the weekly trending chart and deleted one month after that subject leaves it.

Deletion covers the stored files, the result rows and the cached thumbnails of flagged content. Aggregate counts that identify nobody may be kept.

If you uploaded reference material and want it removed sooner, or if you are the subject of a scan, the owner of a flagged account, or a representative of either, write to contact@imagevault.ai with the results link. Removal of an entry from a results page is handled by a person, not an automated process.

The service runs on Cloudflare. Discovery queries public platforms directly and through Apify. Automated screening and adjudication use Cloudflare Workers AI and Anthropic. None of them receive your IP address alongside your upload, and none of them are permitted to use your uploads to train models.