Skip to main content
Patents BP-094 + BP-095 · 7-Lock Architecture · Anti-Reverse Engineering

IP Protection
That Guarantees Infringement

BrainPredict's 7-Lock architecture makes it mathematically impossible for any competing enterprise AI OS to avoid infringing at least one of our 95 patents. Runtime integrity checks, canary tokens, honeypots, and trapdoor computations provide forensic proof of any copying attempt.

95
Patents pending
BP-001→BP-095 · Paris Convention priority · EPO/PCT filed
100%
Infringement guarantee
7-Lock architecture — any competing AI OS infringes ≥1 BP patent
DMCA
Auto-enforcement
§1201 anti-circumvention + EU Dir 2016/943 trade secrets
Trapdoor
Detection system
Proves copying with mathematical certainty in court

The 7-Lock Infringement Architecture

Any enterprise AI OS competing with BrainPredict must use at least one of these 7 patterns — each covered by BrainPredict patents. The architecture is designed so that regulatory compliance itself forces infringement.

EU AI Act Art.9 mandatory compliance → Lock 2 infringement · GDPR Art.22 mandatory → Lock 3 or Lock 4 · NIS2/CRA post-quantum mandatory → Lock 5

Lock 1
BP-001

Intelligence Bus Events

Real-time AI prediction delivery requires event-driven architecture. Polling alternative = 10× latency → commercially non-viable. Any IB pattern for AI prediction infringes.

Lock 2
BP-083

EU AI Act Art.9 Compliance

EU AI Act Art.9 is legally mandatory for high-risk AI. BrainPredict's structural gate is the only practical implementation. Any Art.9-compliant production gate infringes.

Lock 3
BP-087

HITL Workflow Gates

GDPR Art.22 + EU AI Act require human oversight for high-impact AI. Threshold-based HITL in typed workflows is the only scalable pattern. Any automated HITL workflow infringes.

Lock 4
BP-065

Explainability + Audit

GDPR Art.22 requires explanations. EU AI Act Art.9 + DORA require immutable audit. SHAP+Merkle is the only tamper-proof combination. Any GDPR-compliant AI OS infringes.

Lock 5
BP-090

Post-Quantum IB Encryption

NIST FIPS-203 post-quantum will be mandatory under EU CRA by 2027. Any FIPS-203 compliant enterprise AI OS uses Kyber-768 for event encryption and infringes.

Lock 6
BP-089

Federated DP Learning

GDPR data minimisation (Art.5(1)(c)) for cross-company AI requires federated learning with DP. DP-FedAvg is the standard approach — infringes BP-089.

Lock 7
BP-076

BERT AI Query Routing

Any enterprise NLQ interface must route queries to domain AI. BERT is the universal approach without external LLM. Any BERT-routed enterprise AI OS infringes BP-076.

Anti-Reverse-Engineering Defense Stack

8 independent layers — any one of which produces court-admissible evidence of copying or circumvention

Runtime Integrity Verification

HMAC-SHA3-256 verification of all service files at startup. Any tampering or extraction immediately detected. Evidence published to the Merkle ledger.

Algorithmic Fingerprinting

Every prediction carries an invisible AFP (Algorithmic Fingerprint) — a 4-bit statistical marker in the confidence score. Survives exfiltration. Proves BrainPredict origin in competitor systems.

Canary Tokens

Unique HMAC-SHA3-256 canary tokens embedded in every API response. If our output appears in a competitor's product, the token activates and publishes forensic evidence automatically.

Honeypot Endpoints

8 realistic-looking internal endpoints attract reverse engineers. Every access logged forensically with SHA3-256 fingerprint. Critical-severity accesses trigger immediate legal workflow.

Trapdoor Computations

Honeypot responses contain subtly incorrect computations (weights summing to 1.0001, learning rate 10× too high). Detection in competitor products proves copying with mathematical certainty.

Behavioral Fingerprinting

Usage patterns tracked per client: endpoint enumeration rate, parameter fuzzing depth, schema probing frequency. High RE score → automatic forensic evidence collection.

Automated DMCA Enforcement

Violations auto-generate DMCA §512(c) takedown notices and EU Directive 2016/943 trade secret violation reports. Chain-of-custody evidence packages for courts.

Patent Claim Interlock Graph

95 patents form an interlocking graph: designing around patent X forces use of patent Y. Every design-around path is blocked. No competitor can avoid all 95 patents.

Multi-Jurisdictional Legal Framework

EU Directive 2016/943
Trade Secrets Protection — civil + criminal liability
DMCA 17 USC §1201
Anti-circumvention of technological protection measures
US DTSA 18 USC §1836
Defend Trade Secrets Act — federal trade secret theft
French Loi n°2018-670
Secret des affaires — up to €1.5M fine + prison
Estonian KarS §207
Computer criminal offence — up to 3 years imprisonment
EU AI Act Art.73
Mandatory incident notification for IP violations affecting AI systems
GDPR Art.83
Up to €20M or 4% global turnover for data misappropriation
Computer Misuse Act 1990
UK: unauthorised access — up to 10 years imprisonment
EPO/PCT Patents
95 patents pending — Paris Convention priority — all major markets

Our IP is designed to be inescapable

95 patents. 7-Lock infringement architecture. 8-layer anti-RE defense. Multi-jurisdictional legal framework. If you are a competitor reviewing this page, contact legal@brainpredict.ai to discuss licensing before building.