Three tiers.
Five dimensions. One ship gate.
No Hive IP Intelligence product ships public below Gold (85). Every product is scored on five dimensions, 20 points each: Data Completeness, Data Accuracy, Data Freshness, Analytical Depth, Integration Ease. The tier determines what the deliverable can be used for — internal research, client-facing memo, or court-filed exhibit.
What each grade unlocks.
A higher score doesn't mean "fancier." It means the deliverable is engineered to withstand a more adversarial review — your partner reading your draft, your client reading your memo, an opposing expert reading your exhibit.
Gold
· Watermarked "INTERNAL — NOT FOR CLIENT DISTRIBUTION"
· No external reliance
· Standard EULA only
Platinum
· Attorney signoff required
· Kovel letter + BAA executed
· Watermarked "ATTORNEY-REVIEWED DELIVERABLE"
· Reliance scoped to issuing attorney
Double Platinum
· Attorney signoff required
· Chain-of-custody receipt (on-chain DID + PQ sig)
· Verifiable compute attestation (model + input + output hashes)
· FRE 901/902 package included
· Audit-trail export available
Where every product stands today.
"Today" is the live score that ships now. "This week" is the engineering target by Friday. "Stretch" is the free-data ceiling for the product, with any optional paid bridge noted. Six products live today. Two (HiveLicense, HiveDesign) are in research preview with a partner and will publish scores once they clear the Gold (85) gate.
How every product is scored.
Each dimension is worth 20 points. The composite is the sum.
Data completeness
Percentage of required data fields covered by the underlying sources. Missing fields are scored at zero.
Data accuracy
Percentage of data that is correct against ground truth (authoritative government register or court filing).
Data freshness
How current the data is. Real-time is 20; daily is 16; weekly is 12; monthly is 8; stale (>30d) is 0.
Analytical depth
Richness of the analytics layered on top of the raw data — examiner profiles, judge analytics, citation networks, SEP overlays, FRAND models.
Integration ease
API quality, documentation, reliability, rate-limit headroom for our workload. Free government APIs typically score 14–18 here.
Why we don't lead with a SOC 2 badge.
Hive's trust architecture — post-quantum cryptographic receipts, on-chain audit trail via deterministic DIDs, Kovel-extended attorney-client privilege, and verifiable compute attestations — satisfies every SOC 2 Type II Trust Services Criterion and adds trust dimensions SOC 2 has no controls for.
The AICPA framework hasn't written controls for what we built. Cryptographic non-repudiation of AI inference. Privilege chain provenance. Quantum-resistant data sealing. None of these are SOC 2 criteria. The Clarity Act movement and emerging AI accountability legislation make this posture affirmatively defensible.
We are not replacing SOC 2 with nothing. We are replacing it with a higher bar SOC 2 cannot express. The Double Platinum tier is the public proof point: every receipt at that grade is FRE 901/902 self-authenticating without an audit firm signing off.
A SOC 2 Type II audit ($60–120K via Prescient Assurance or equivalent) is available on request for enterprise channel credentialing where customer procurement processes require it. It is not gating product launch.
What this actually costs to run.
Recurring data spend for the full eight-product slate is $0 – $12,000/year. Most products run on free government and nonprofit APIs (USPTO, EPO, SEC EDGAR, CourtListener, GLEIF, OFAC, EUR-Lex CELLAR, WIPO, Federal Register). Optional paid bridges total $12K/yr if every product reaches its stretch score.
Reference: comparable commercial-data subscription bundles (Clarivate Derwent, Questel Orbit, Lex Machina, Darts-IP, ktMINE, Reed Tech) total $230,000 – $655,000 / year. Hive's free-first architecture replaces them with government APIs and an LLM synthesis layer at 0.5% of the cost for ~95% of the analytical capability.
How to verify any Hive deliverable.
Every Hive deliverable carries an on-chain receipt with the deterministic DID of the issuing agent, the ML-DSA-65 post-quantum signature of the output, and the hash of every input artifact used. At Double Platinum tier the verifiable compute attestation is included as well — the exact model version and prompt that produced the output.