The classification corpus
Every layer is legally relevant.
Every layer must be evaluated.
The HTSUS is not a flat list of headings. It is a layered legal instrument — eleven distinct levels of authority, each governing a different aspect of classification. Add the CBP regulatory corpus and you have fifteen layers that a complete classification must account for.
Why the regulatory corpus matters
A classification that is correct under the HTSUS but contradicts an active CBP ruling on substantially identical merchandise is a classification that will not hold. The regulatory corpus is not supplementary material — it is authority.
Sequence is not optional
The right corpus
and the right order.
Having access to the corpus is necessary but not sufficient. Each layer governs at a different level of specificity, and the precedence relationships between them are legally defined. A tool that evaluates heading text before screening chapter notes has applied the law backwards. A classification that identifies a correct heading but fails to check applicable Chapter 99 overlays is an incomplete entry.
Kanon applies each layer in the sequence CBP applies them — and records not just the outcome but which layer was determinative at each step. This is what allows a classification to be reconstructed, explained, and defended.
User-in-the-loop classification
No black boxes.
Every decision point is explicit.
Kanon does not produce a code. It produces a classification — a chain of explicit determinations, each grounded in a specific layer of the corpus, each visible to the user who is responsible for the entry.
At each decision point, the user can see the provision being applied, the product attribute it is being applied to, and the outcome. Factual determinations that the engine derived — rather than received from explicit product data — are flagged so the user can confirm or correct them. Confirmed determinations are recorded with the user's identity and timestamp.
This is what makes the classification auditable. Not a logged output — a transparent process in which the human expert remains in control at every determinative step.
Legally substantiated
Every determination
documented at the moment it is made.
The Classification Support Package Kanon generates at the moment of classification is an immutable record. It captures every corpus layer evaluated, the outcome of each evaluation, the product attributes driving factual determinations, the HTSUS corpus version in effect, and the identity of every user who confirmed a determination.
This record is generated when the classification is made — not reconstructed when a CBP inquiry arrives. The distinction matters: a retroactively assembled rationale is not the same as contemporaneous documentation. Kanon produces the latter.
How Kanon compares
Corpus coverage is where
most tools stop short.
| Corpus layer evaluated | General-purpose AI | HTSUS lookup tools | Kanon |
|---|---|---|---|
| GRI (all six, in sequence) | ✗ | ✗ | ✓ |
| Additional U.S. Rules of Interpretation | ✗ | ✗ | ✓ |
| General Notes | ✗ Partial / inconsistent | ✗ | ✓ |
| Section Notes | ✗ Partial / inconsistent | ✗ | ✓ |
| Chapter Notes | ✗ Partial / inconsistent | ✗ Manual lookup | ✓ |
| Subheading Notes | ✗ | ✗ | ✓ |
| Additional U.S. Notes | ✗ | ✗ | ✓ |
| Chapter 99 overlay evaluation | ✗ | ✗ | ✓ |
| Statistical suffixes (full 10-digit) | ✗ Often 6-digit only | ✗ Varies | ✓ |
| CBP binding rulings / revocations | ✗ | ✗ | ✓ |
| Immutable CSP with full reasoning trace | ✗ | ✗ | ✓ |
| Pinned to a versioned corpus | ✗ Training data cutoff | ✗ Typically unversioned | ✓ |