CROWN Receipts

CROWN receipts are the cryptographically signed proof artefacts that form the trust foundation of the entire CueCrux platform. Every time Engine generates an answer, it produces a CROWN receipt that immutably records every decision made during the retrieval and synthesis process, creating a verifiable audit trail that anyone can independently inspect and validate.

Each CROWN receipt captures several layers of information. At the top level, it records the original query exactly as it was received, along with the assurance mode that was selected and the timestamp of the request. This establishes the context in which the answer was produced. Below this, the receipt contains the complete list of retrieved chunks with their relevance scores, embedding lane metadata, and source artefact identifiers. This evidence layer shows exactly what information was available to the synthesis model and how it was ranked.

The synthesis rationale section documents how the language model constructed the answer from the available evidence. It records which chunks were cited for each claim, what confidence the model assigned to each assertion, and any caveats or limitations it identified. This rationale makes the answer's reasoning transparent, allowing reviewers to assess whether the conclusion follows logically from the evidence.

Binding all of this together is a BLAKE3 content hash chain. Each element of the receipt is individually hashed, and these hashes are combined into a Merkle-like tree structure that produces a single root hash representing the entire receipt. This root hash is then cryptographically signed using the platform's signing key. The hash chain means that any modification to any part of the receipt, whether to the query, the evidence, the scores, or the rationale, would invalidate the root hash and break the signature. This makes receipts tamper-evident in a mathematically provable way.

Independent verification is a key design goal. Anyone with a CROWN receipt can verify its integrity without relying on the CueCrux platform itself. The verification process checks the hash chain, validates the signature against the platform's published public key, and confirms that the receipt structure conforms to the published schema. SDKCrux provides one-line verification helpers, but the algorithm is simple enough that third parties can implement their own verifiers.

Receipts are designed to be shareable through proof links that can be embedded in reports, regulatory filings, and legal documents. A proof link resolves to a read-only view of the receipt with the full evidence chain visible. This means that any claim backed by a CueCrux answer can be independently verified by anyone with the proof link, even years after the answer was produced.

The receipt system also supports redaction for privacy-sensitive scenarios. When evidence comes from private corpora, VaultCrux can produce redacted receipts that hide the underlying content while preserving the hash commitments. A redacted receipt proves that evidence was considered and contributed to the answer without exposing the confidential source material.

For the CueCrux platform, CROWN receipts are what transform AI-generated answers from unverifiable assertions into auditable, defensible artefacts. They are the mechanism that earns trust in an era of AI-generated content.