The Legal Regime Around Discovery
Competitive intelligence is not constrained by what an analyst is allowed to know. It is constrained by how the analyst came to know it. The governing instruments are uniform on this point. The Defend Trade Secrets Act and the Economic Espionage Act turn on whether information was acquired by improper means. The EU Trade Secrets Directive defines unlawful acquisition by reference to the conduct of collection rather than the value of the result. The Computer Fraud and Abuse Act and the Electronic Communications Privacy Act criminalize particular methods of access. Section 5 of the FTC Act reaches unfair methods of competition. The SCIP Code of Ethics, while not law, is the professional standard against which a practitioner's conduct is measured. Each of these instruments evaluates the manner of discovery, and each presumes that the manner of discovery can be reconstructed and examined.
The prevailing tooling cannot satisfy that presumption. Keyword crawlers, alerting services, and retrieval-augmented language model assistants return findings without retaining a defensible record of the collection path that produced them. When a competitor alleges misappropriation, the intelligence function is asked to prove a negative: that no protected source was touched, that no access control was circumvented, that every signal traces to a public disclosure. A pile of results does not prove this. The architectural requirement that the legal regime imposes is a search process whose own history is a first-class, governed, auditable artifact.
What Conventional Intelligence Tooling Lacks
A conventional query is stateless. It is evaluated once against an index and discarded; it carries no persistent record of the competitive question it was serving, no accumulated memory of what prior steps already established, and no governance evaluation of the transitions it made along the way. Retrieval-augmented generation architectures compound the problem: a separate search engine retrieves documents, a separate language model synthesizes them, and semantic context is lost at every interface crossing. There is no persistent query entity and no admissibility check on intermediate retrieval or reasoning steps. The system can tell you what it concluded. It cannot show you, step by step, how it got there or prove that every step was permitted.
This is precisely the gap that turns a routine intelligence exercise into legal exposure. Without a recorded, policy-governed collection path, the difference between lawful open-source analysis and unlawful acquisition collapses into the analyst's unverifiable account of their own process.
What Semantic Discovery Provides
Semantic discovery, disclosed in United States Patent Application 19/647,395, instantiates each competitive question as a discovery object: a persistent, memory-resident semantic entity that carries the full context of the inquiry as a typed data structure and traverses a governed index of sources anchor by anchor. The discovery object is not a keyword list or a prompt. It comprises typed fields disclosed in the application, including an intent field encoding the strategic objective of the inquiry, a context block encoding domain and temporal and privacy parameters, a memory field accumulating the partial findings established by prior admitted steps, a policy reference field carrying the governance constraints that apply to the collection, and a lineage field recording the ordered sequence of admitted transitions.
At each anchor boundary, the discovery object undergoes a single three-in-one traversal step in which search, inference, and governance fuse rather than running as separate subsystems. The step simultaneously narrows the search space across the available sources, updates the discovery object's semantic state, and evaluates the proposed next transition for admissibility under deterministic policy. Governance is a constituent phase of the step, not a filter bolted on afterward. As disclosed in the application, the model proposes and the substrate decides: an inference component proposes the next transition, and the governance component at the anchor independently evaluates it against policy constraints, lineage continuity, and entropy bounds, producing one of three outcomes. An admitted transition advances the traversal. A rejected transition is blocked and the traversal backtracks. A decomposed transition is broken into finer sub-transitions that are individually re-evaluated. This structure lets an organization put an untrusted or highly capable language model in the proposing role without surrendering control of what the collection process is permitted to do.
The lineage field is what answers the legal question. It records, for every admitted transition, the anchor identifier, the timestamp, the semantic state mutation that was applied, the admissibility determination that permitted the transition, and the neighborhood from which the transition was selected. The application further discloses that the governance history records not only admitted transitions but the transitions that were evaluated and rejected, the reasons for rejection, and the decomposition paths taken. The result is a complete, reproducible account of how each strategic signal was found and a documented record that the collection never entered a source policy forbade. As disclosed, the admissibility evaluation is deterministic: the same discovery object encountering the same proposed transition produces the same outcome, so the collection path can be independently re-verified rather than merely asserted.
The per-step admissibility overhead is bounded. The application discloses that admissibility evaluation operates on the bounded, actively maintained semantic neighborhood of the current anchor and the anchor's own governance configuration, which is what makes per-step governance practical across a large traversal rather than a prohibitive cost.
Embodiments and Deployment Options
The substrate admits several deployment configurations for a competitive intelligence function.
Source scoping by policy. The policy reference field can encode source-class constraints so that the discovery object is structurally barred from neighborhoods representing non-public or access-controlled material. A traversal configured for public-disclosure-only collection will see transitions into restricted neighborhoods rejected at the anchor and recorded as rejections, producing affirmative evidence of restraint rather than the absence of evidence.
Heterogeneous source federation. Because every addressable source object is reached through the same governed traversal and the same structured alias addressing, a single discovery object can follow a strategic concept across patent corpora, regulatory filings, hiring signals, and earnings disclosures without losing semantic context at the boundary between source types. Alias resolution is navigational, resolving an address by traversing the live index rather than by lookup against a separate table, so a competitor's renamed or reorganized public footprint remains resolvable through the alias resolution protocol without breaking the collection path.
Persistent monitoring versus one-shot inquiry. A discovery object can be instantiated for a single freedom-to-operate or competitive-positioning question and terminated on resolution, or it can persist across cycles to track an evolving competitor program, accumulating memory across runs while every incremental step is admitted and recorded under the same governance regime.
Modulated traversal posture. The discovery object's affective and confidence fields, disclosed in the application, parameterize how aggressively a given traversal explores. A novelty-weighted posture favors uncharted semantic territory when the objective is to surface non-obvious strategic moves; a more conservative posture favors well-supported transitions when the objective is defensible confirmation. The same competitive question can be run under different postures to produce different but individually auditable collection paths.
Untrusted proposer integration. Because governance is independent of the proposing component, an organization can adopt a third-party or frontier language model to propose candidate transitions while the admissibility substrate, running under the organization's own signed policy, retains decisive authority over what the collection is allowed to do.
Why This Is the Compliant Architecture
Against the trade-secret regime, the substrate produces a per-traversal record that documents the manner of acquisition, the precise fact the Defend Trade Secrets Act, the Economic Espionage Act, and the EU Trade Secrets Directive make dispositive. Against the Computer Fraud and Abuse Act and the Electronic Communications Privacy Act, policy-scoped admission means access methods the organization has not authorized are blocked at the anchor and the block is recorded. Against Section 5 of the FTC Act and the SCIP Code of Ethics, the lineage is the documentary basis for demonstrating that the collection followed declared, lawful methods. In each case the compliance artifact is generated as a byproduct of operation rather than reconstructed under litigation pressure after the fact.
The strategic implication is that competitive intelligence moves from a practice defended by the analyst's account of their own conduct to a practice defended by a deterministic, reproducible record of the collection path. The substrate does not replace the analytic judgment of the intelligence function. It supplies the governed, auditable discovery layer that the legal regime around competitive intelligence has always assumed and that conventional search tooling has never provided.
Disclosure Scope
This article describes an application of the semantic discovery invention disclosed in United States Patent Application 19/647,395 to the domain of competitive intelligence. The competitive intelligence domain framing, the legal-regime analysis, and the deployment scenarios are application context. The underlying technology, including the discovery object and its typed fields, the three-in-one traversal step fusing search, inference, and governance at each anchor boundary, the admit, reject, and decompose admissibility outcomes, the per-traversal lineage, deterministic and bounded-overhead admissibility evaluation, and navigational alias resolution, is disclosed in United States Patent Application 19/647,395. This article is an enabling public disclosure of that application and does not introduce ranking metrics, benchmark figures, or mechanisms beyond those disclosed in the cited application.
