The trust layer around your DRM.
A multi-cloud content-security and executable-rights platform for premium broadcast and streaming. It runs alongside Widevine, PlayReady, and FairPlay — not instead of them — adding a cryptographically independent control plane for attestation, executable rights, live anti-piracy, and provable settlement. No catalogue migration. No new DRM protocol.
Authority is a consequence, not a secret.
Cryptographic authority is derived from verified state, executable policy, and authorized rights — never distributed as an unconditional static secret. A playback or media operation succeeds only when the requesting party can prove every condition the policy requires: entitlement, hardware and output protection, release window, territory, risk state, watermark obligation, custodian quorum. Relay then records enough evidence to prove afterward what was decided, why, and under which security state.
Your DRM keeps its job. Relay adds the one it can't do.
Widevine, PlayReady, and FairPlay keep handling entitlement, key delivery, and protected playback in their own ecosystems. Relay wraps an independent control plane around them — so a compromise of an API node, a policy service, or a single cloud account can't reconstruct root authority or forge a valid media decision.
No migration
CDNs keep distributing conventionally packaged encrypted media, and existing entitlement systems stay authoritative for subscriptions. The security layer carries no video payload — a 50 Mbps UHD stream does not create a 50 Mbps Relay workload.
The Trust Gateway
Sits logically adjacent to the existing licence path: rights resolution, attestation, risk, and watermark allocation, then hands off to the DRM licence proxy. Existing DRM keeps the platform-specific protocol.
Fail-closed by design
A cryptographic-policy failure never silently becomes unrestricted access. High-value live uses fail-secure semantics; degraded-operation rules exist only where a rights owner explicitly permits them.
Contracts that compile.
The Rights Compiler turns distribution agreements, territory matrices, sports and blackout schedules, embargoes, talent restrictions, and AI-usage terms into a canonical, signed representation — one that compiles to multiple enforcement targets at once.
Delegation that can only narrow
Rights are attenuable capabilities. A distributor or broadcaster can consume, delegate, or narrow what it received — never expand it. No downstream party can invent a new territory, a longer window, a higher resolution, or broader AI permission.
Inheritance through derivatives
Localized audio, trailers, clips, accessibility assets, and AI-generated derivatives inherit the restrictions of their source. “Training prohibited” stays attached down the chain, joined to provenance through C2PA.
Fail-closed compilation
If an enforcement target can’t represent a mandatory security property, compilation fails rather than silently weakening the contract. A legacy device that can’t measure its runtime simply isn’t issued 4K.
Trust derived from measured hardware.
The Media Attestation Service normalizes heterogeneous evidence — AWS Nitro, AMD SEV-SNP, Intel TDX, Google Confidential Space, NVIDIA confidential GPUs, OCI SEV-SNP, OEM TEEs, STB and Smart TV trust anchors — into one profile. Verification is nonce-bound against replay, with a policy-specific freshness window. Profiles don't replace Widevine L1/L2/L3; they give rights owners one policy language across every ecosystem.
| Profile | State | Eligible content |
|---|---|---|
| MEDIA-0 | Untrusted execution | Premium unavailable |
| MEDIA-1 | Software-secured | Limited resolution, high-frequency renewal |
| MEDIA-2 | Verified device posture | HD, per rights-owner policy |
| MEDIA-3 | Hardware-protected execution + secure decode | Premium HD / UHD |
| MEDIA-4 | Fresh attestation + approved measurements + protected output | Highest-value UHD / live |
The stream that was ripped is the stream that goes dark.
Premium live runs on continuously renewable authority: the stream is divided into cryptographic epochs, and each continuation requires the session to still be policy-valid — account, entitlement, device trust, risk state, and known piracy evidence, re-checked every epoch. When a pirate restream is detected and its watermark recovered, the source session is identified, an abuse proof is validated, and the next epoch is refused.
Forensic watermark binding
Each protected playback carries a watermark derived from session, content, and policy. A recovered watermark resolves to the offending session — without exposing subscriber identity to unrelated parties.
Content Abuse Proof
A detected pirate source produces a signed, portable proof that a validated abuse event occurred. Subscriber data stays under the provider’s control; the proof is what travels.
Collective Defence Fabric
Providers exchange abuse proofs, not a mandatory global blacklist. Each applies its own response — monitor, force re-auth, elevate watermarking, shorten renewal, downgrade, or refuse — over scoped pseudonyms.
Relay doesn't claim to stop a camera pointed at a screen. It closes the loop between forensic detection and the next authorized frame.
Prove what happened. Prove what didn't.
Every security-sensitive decision produces an immutable commitment; the raw event stays encrypted and access-controlled, while authorized parties request selective cryptographic proofs. This is materially different from an audit trail, which only reports what an application claims it did.
Proof of Rights Enforcement
Each decision binds the signed policy, the rights capability, the execution state, and the outcome — so a verifier can establish that the decision actually conformed to policy, not that a log says it did.
Proof of Non-Issuance
Bounded negative proofs: that no pre-release licence issued before the contractual timestamp, that no generative-AI operation was authorized against a catalogue in a period, that no UHD licence issued below a required profile.
Proof-to-Pay
The same signed execution events that enforce rights aggregate into verifiable usage statements — per playback, stream-minute, live event, AI inference, or training job. Enforcement and settlement refer to one record.
No single cloud holds the keys.
The control plane runs active-active across AWS, GCP, and OCI. For high-value content, authority is split with threshold custody — a 2-of-3 across a studio HSM, an SCSA HSM, and a broadcaster HSM, or shares spread across clouds with one held in the rights owner's own infrastructure. Loss or compromise of any single cloud provides no unilateral authority, and administrative override is itself a policy-controlled, multi-party, logged operation.
Machines are principals too: localization, transcription, indexing, approved inference, and contractually authorized training run inside confidential-compute environments that must attest before an encrypted asset is released. Relay proves a narrow, checkable statement — that protected content was released only to a specific measured workload, running an approved operation, under a defined policy.
Relay doesn't represent analogue capture as cryptographically impossible — a camera can record displayed pixels, and post-decode escape is addressed through attribution, forensic evidence, and revocation, not by pretending it can't happen. Cryptography here proves integrity, provenance, and that defined conditions were met; it doesn't prove a prediction about the future is true. The platform's output isn't a key — it's a verifiable answer to who or what was authorized to perform this operation, on this asset, under which rights, against which measured state, for how long, and what evidence proves the decision was enforced correctly.
Built on the seven-layer SCSA core. See the platform →