Most enforcement technology in the IP space targets one step in the process. Detection tools find infringement. Demand automation generates claim letters. Litigation analytics predict outcomes. Collection tools manage post-judgment recovery.
Copydefend operates all four functions inside one infrastructure. The architecture carries a case from automated detection through pre-trial demand, court filing, judgment, and post-judgment recovery within one operational dataset.
The Three-Layer Architecture
The pipeline splits into three connected layers. Detection and portfolio construction sit at the top. Risk-assessed litigation in the middle. Collection optimization at the bottom. Each layer produces operational data that feeds the others.
Layer One. Claims Portfolio
Detection runs through perceptual image hashing combined with proprietary e-commerce image search trained on marketplace product card data. Authorship verification operates as a two-tier system.
Portfolio construction sorts cases through two-dimensional author categorization, combining economic density with licensing risk profile. A compliance gate enforces a 10-day delay between detection and demand. This is a defense against false positives, not a procedural delay.
Layer Two. Risk Litigation
Pre-filing assessment runs through a probabilistic expected revenue model that combines win likelihood, expected compensation, and projected timeline. Concentration limits per court and per author defend against systemic risk.
The component worth pointing out specifically is judge-level conditional expected value modeling. Standard litigation analytics work on court-level averages, which compress the signal. For copyright cases, where individual judges within the same court have measurably different award patterns, the average obscures more than it reveals.
The Copydefend approach segments active judges within a court into sub-distributions and calculates expected value as a weighted sum across those sub-distributions, not as a single point estimate. In academic legal analytics this is called conditional outcome modeling. It appears rarely in production systems.
Layer Three. Collection Optimization
Post-judgment work runs through the Russian Federal Bailiff Service. The platform tracks three signals: voluntary versus enforced payment attribution, bailiff office performance disaggregated by region and case type, and the gap between high win rates and actual cash recovery.
The third signal matters most. A 98.5% court win rate is a different number from actual cash recovered. The collection layer measures the difference and feeds it back into earlier stages of the pipeline.
The Closed Loop
Operating all three layers inside one architecture produces a closed calibration loop. Judgment outcome data calibrates the risk litigation model. Recovery data calibrates portfolio construction. Pre-trial settlement data calibrates detection thresholds. Each layer feeds the others.
Single-function tools cannot run this loop. A detection vendor sells leads. A claims automation tool sells letters. A litigation analytics platform sells predictions. None of them sees the full cycle, so none of them can calibrate based on actual outcomes.
Vertical vs Horizontal Infrastructure
In the 2026 LegalTech landscape, horizontal AI for the legal profession occupies one category. Vertical enforcement architecture with closed operational loops occupies another. The technology stacks overlap. The customer models do not.
A wrapper over frontier LLMs can be built in two weeks. A vertical pipeline with judicial scoring trained on real outcomes, image search tuned on actual marketplace data, and bailiff recovery analytics structured from years of execution proceedings takes years to build and longer to refine.
The asset is operational data combined with vertical workflow design. Both compound over time. Neither can be bought.