Disclosure tells you what an uploader claims. Detection tells you what your model guesses. Neither is a record a regulator, a rights-holder, or a court can verify without trusting the platform. As Article 50 of the EU AI Act turns transparency from policy into law, the question stops being “did you label it” and becomes “can you prove the label was true, and that no one changed it.”
The 2026 stack is real and moving fast. But every layer of it produces a claim that the claiming party controls. That is exactly the property that fails when the claim is contested.
Spotify’s AI Credits and Apple’s Transparency Tags use the DDEX standard. An uploader ticks a box saying how AI was used. It is honest until it isn’t. A self-declaration is a claim by the party with the most incentive to shade it.
Deezer auto-tags fully-AI tracks. That is reportedly 44% of its daily uploads, about 75,000 songs a day. Detection is probabilistic, operator-owned, and contestable. “Your model guessed” is not the same as “the creator proved.”
C2PA credentials and SynthID-style watermarks are valuable, but metadata can be edited and watermarks can be laundered through re-encoding. A label that can be removed is not a record that can be relied upon.
Hive signs the asset together with its declared provenance: human, AI-assisted, or fully AI, the model, the rights basis, and the timestamp, using the named profile. A relying party can recompute the profile's checks with its published key and compatible verifier. Hive operates the signer, so institutional independence requires non-Hive key and log custody. It does not replace DDEX, C2PA, or your detector.
On June 10, 2026 the European Commission and the AI Office published the final Code of Practice on marking AI-generated content. On August 2, 2026, Article 50’s machine-readable marking obligation becomes enforceable, with penalties reaching 3% of global turnover. The platforms that distribute synthetic audio and video are squarely in scope.
A label can be disputed. A signed receipt makes the declared fields tamper-evident under the named key, but it does not make those declarations true or create institutional independence.
Hive is not another labeling scheme competing with C2PA or SynthID. It is the evidentiary layer that makes whatever label you apply hold up when an artist, a rights-holder, or a regulator asks you to prove it.
When a track or video is delivered, Hive signs a receipt binding the asset hash, the declared provenance fields (human / AI-assisted / fully AI, model, rights), and the timestamp. This happens out of band. It records, it does not gate ingestion.
The receipt rides with the existing metadata. Your disclosure and detection pipelines are unchanged; the signature simply makes their output independently verifiable.
An artist disputes a label, a rights-holder alleges a voice clone, a regulator audits. The receipt verifies, or it doesn’t, with the public key alone, and no access to your systems. You move from “trust our records” to “check the math.”
Verification costs nothing and needs nothing from Hive. The receipt outlives the dispute, the catalog migration, and Hive itself.
This is a Hive-operated signer, not a mock. Pick your world, then your exact channel. You take the steps: sign the declared fields, check them with the matching verifier, then relabel one field and watch verification fail.
1 · Pick your world
2 · Pick your channel
Streaming, video, social, advertising, live and broadcast, messaging, official data feeds, AI output, and human sign-offs can use a configured provenance profile. Verification depends on the named schema, key, verifier, and retained resources. Hive can sit underneath an existing disclosure or detection workflow.
Three steps, the same for every channel. You take them; these are the expected results. Hive operates the countersigner and log, so the evidence is independently checkable. True institutional independence requires non-Hive key and log custody. It makes no decision about the content. It signs what you declare and lets anyone else check it.
Send the item and its declared provenance (what it is, which model, the rights, the time) to the signer through one API call from your channel.
result: typed fragments acceptedThe selected profile produces an ML-DSA-65 receipt over the declared fields using a Hive-operated key that is separate from the submitting platform's key. This does not establish neutral-party independence.
result: Hive-operated receipt in ~77µsA holder can check the signature with the matching verifier, trust key, receipt, and retained resources. The result establishes integrity under that profile, not the truth of every declared field.
result: VALID, or INVALID if alteredThe signature comes from a Hive-operated key rather than the platform operator's key. That separation is not neutral-party certification or institutional independence.
Hive attests only to what was declared and that it was unchanged. It does not assess whether the content is good, true, or allowed.
Runs beneath the DDEX, C2PA, SynthID, detection, and governance you already use. It does not replace or compete with any of them.
Pick your world, then the channel you publish to. You will land on a page that sets up your tenant and hands you a one-line call to sign your first receipt. No call, no demo, no one to talk to.
This is Hive-operated provenance and attestation that is independently checkable. Hive signs what you declare and makes no decision about your content. You run the steps above; the expected results are VALID on an honest receipt and INVALID the instant a field is altered. Verification is free for anyone.
No call, no demo, no one to talk to. Sign a receipt above, verify it independently with the public key alone, then change one field and watch it fail. The same flow runs against your own channel through the SDK. It is Hive-operated provenance that is independently checkable and sits underneath the DDEX, C2PA, detection, and governance you already run.
Sources: EU AI Act Article 50 enforceable Aug 2, 2026, fines to 3% of turnover: Chartlex, sota.io, Mintec. Final Code of Practice published Jun 10, 2026: ComplianceHub, Licentium. Platform policies (Spotify/Apple DDEX, Deezer detection): Chartlex, Deezer. YouTube auto-labels and C2PA: NerdLevel Tech.