Local first
Your private keys, trace secrets and original master stay on your device.
notdataCreator rights & traceability infrastructure
Publishing should not mean surrendering your work.
NOTDATA is open-source infrastructure to prove what you created, declare AI-training rights, attach privacy-preserving trace identities and progressively validate stronger media-specific defenses and training-use audits.
Portable proof
Evidence and Trace-ID descriptors can be verified without a NOTDATA account or hosted portal.
Honest protection
NOTDATA separates provenance, rights, trace identity, training-use detection and anti-training instead of pretending they are the same guarantee.
Protection model
Know what your protection actually means.
NOTDATA uses explicit levels instead of a vague protection score. Higher levels require stronger, media-specific evidence; unknown or untested capabilities stay visible.
P0UNPROTECTED
No NOTDATA protection evidence is available for this artifact.
Claim boundary: P0 makes no provenance, rights, binding, anti-training or traceability claim.
P1PROVENANCE
Signed provenance and artifact commitment can be independently verified.
Requires
- • signed provenance
- • artifact commitment
- • portable evidence
Claim boundary: P1 does not claim active resistance to copying or AI training.
P2RIGHTS
Verifiable machine-readable AI-training rights are attached to authenticated provenance.
Requires
- • P1
- • signed AI-training rights
- • verifiable rights history
Claim boundary: P2 makes permission state explicit; it does not technically prevent copying or training.
P3BOUND
The published artifact is durably linked to its signed provenance and rights evidence.
Requires
- • P2
- • media-specific durable binding
- • applicable robustness evidence
Claim boundary: P3 is only justified for transformations actually covered by the cited benchmark evidence.
P4DEFENDED
An active defense has passed the applicable NOTDATA benchmark for this media/profile.
Requires
- • P3
- • active defense
- • human-utility measurement
- • measured training-utility evidence
Claim boundary: P4 must not be inferred from watermark survival alone; Training Utility Gap must be measured where claimed.
P5TRACEABLE
Validated training-use tracing is available for the supported scenario in addition to lower-level protections.
Requires
- • P4
- • validated trace carrier
- • calibrated training-use detection
- • applicable removal-attack evidence
Claim boundary: A valid Trace-ID alone never establishes P5 or proves that a model trained on the work.
Inspect a Protection Report locally
Paste protection-report.json and its evidence.json. NOTDATA verifies the report commitment, its exact evidence binding and artifact binding in your browser. Nothing is uploaded.
Verify evidence locally
Paste a NOTDATA evidence bundle. Nothing is uploaded; verification runs in your browser.
Selecting the artifact also verifies its signed SHA-256 commitment. The file remains local to this browser session.
Verify a Trace-ID locally
Verify that a signed Trace-ID belongs to exact NOTDATA evidence. You may optionally select your local 32-byte campaign secret to prove that it derives this artifact Trace-ID. Evidence and secrets stay in this browser session.
The secret must be the raw 32-byte NOTDATA campaign secret. It is never required to verify the public descriptor and is not uploaded by this page.
A verified Trace-ID proves descriptor integrity and its binding to signed evidence. It does not prove that an AI model trained on the work. Artifact and training-use detection remain NOT_TESTED until a carrier and audit method pass their benchmark gates.