What is Sourcemark?

AI companies are training models on creative work — photographs, illustrations, documents, video — often without the knowledge or consent of the people who made it. There is no standard way for a creator to say “yes, you can use this” or “no, you cannot.” Copyright law is slow. Technical measures like watermarking are easy to circumvent and don't scale.

Sourcemark fixes the consent layer. It gives creators a simple way to declare whether their work is available for AI training, and gives AI companies a registry to check before they train.

What Sourcemark does

Sourcemark creates a timestamped, machine-readable recordof your declaration. When you sourcemark a file, the system records:

Who made the declaration (you)
What file it applies to (identified by fingerprint)
When the declaration was made
What you said — whether the work is available for AI training, and under what terms

That record goes into a public registry that AI companies can query programmatically. Think of it as a consent layer between creators and the AI training pipeline.

What Sourcemark does not do

It is important to be clear about boundaries:

It does not enforce copyright.Sourcemark records your declaration. It does not file takedowns or pursue legal claims on your behalf.
It does not prevent unauthorised use.No tool can prevent a determined actor from downloading publicly available files. Sourcemark works at the consent layer, not the access layer.
It does not prove ownership.Your record states that you made a declaration about a file. It does not prove you created it.

Note

Sourcemark is a registry, not a lock. It creates evidence of what was declared and when — not a barrier to access.

How it works

When you add a file to Sourcemark, the file never leaves your device. Instead, Sourcemark generates two types of fingerprints locally:

Exact fingerprint (SHA-256):A cryptographic hash that uniquely identifies the file byte-for-byte. Change a single pixel and the fingerprint changes.
Perceptual fingerprint:A visual similarity hash that can match files even after resizing, recompression, or minor edits. This is how Sourcemark catches near-duplicates.

Only the fingerprints are stored — never the file itself. Your work stays on your machine. The registry holds just enough information to answer the question: “Has anyone made a consent declaration about this file?”

Supported file types

Sourcemark works with images (JPEG, PNG, WebP, TIFF), PDFs, and video files. If your creative work lives in these formats, it can be sourcemarked.

Who it's for

Individual creators— photographers, illustrators, designers, writers, filmmakers — who want a clear, public record of their AI training preferences. Sourcemark is free for individuals.

Studios and agenciesmanaging large libraries of creative work can use paid plans for batch operations, team management, and advanced features.

AI companiesbuilding training datasets can query the registry to check consent status before including a file. Enterprise plans provide API access at scale.

Why it matters

The AI training consent problem is not going away. Models are getting larger, datasets are getting broader, and creators are increasingly aware that their work is being used without any record of permission. Legislation is moving slowly and inconsistently across jurisdictions.

Sourcemark does not wait for the law to catch up. It creates a practical, verifiable standard that works today: creators declare, the registry records, AI companies check. The result is a clear chain of consent that benefits everyone — creators get agency over their work, and AI companies get a clear, documented record of permission.

Ready to get started? Create your first Sourcemark.