Definition
Dual-use poznání (dual-use knowledge) is the structural ethical problem that arises whenever the same understanding is required for both defense and offense in a given domain. The term originates in arms-control and biosecurity literature (chemistry, microbiology, nuclear physics) and has been extended to information security and — through the Kompendium epistemické a kognitivní anihilace — to epistemic security.
The dual-use problem is structurally unavoidable, not a defect of presentation: a clinician who can detect gaslighting understands gaslighting well enough to perform it; a researcher who can document EA-1.3.1 immunization patterns can deploy them; a Red Team operator who can simulate delegitimizace campaigns at scale knows how to run them. Restricting access does not eliminate the knowledge; it only changes who has it.
The kompendium adopts and extends the standard dual-use ethical framework with adaptations specific to epistemic security:
- Asymmetry of damage — defensive use of EA knowledge typically requires shared understanding with the protected subject (clinician + patient, security team + organization). Offensive use is unilateral. The defensive use case structurally favors transparency.
- Public-good knowledge floor — basic recognition of common manipulation patterns is itself a public good. Withholding the recognition vocabulary leaves potential targets without the categories needed to detect what is happening to them. The kompendium argues that not publishing recognition-level material has its own ethical cost.
- Tiered access for technique-level material — specific implementation details, novel attack patterns, and untested techniques warrant restricted access (qualified researchers, clinicians, security personnel). Recognition material is open; technique-novelty material is restricted.
- Proportionality and accountability — practitioners with access accept documentation and oversight responsibilities; access is not anonymous.
Context
Within the Prismatic Platform, dual-use considerations are operationalized through explicit access tiering and audit:
- The Kompendium itself follows the tiered model: high-level taxonomy and recognition material are open at /kompendium/ and /glossary/; per-technique deep-dives at /kompendium/techniques/ include implementation-relevant detail and require authenticated access plus declared professional context.
- The Color Teams Red Team operations — which by design use EA techniques against platform-internal targets — run under audit and are restricted to vetted operators. Red Team scenario libraries are not openly published.
- The Decision Engine records the operational use of any technique-level kompendium reference, providing accountability without restricting legitimate use.
- NABLA Infinity treats EA-related knowledge as a provenance-tagged class of evidence: when EA reasoning informs a decision, the reasoning is recorded explicitly so that later review can distinguish defensive from offensive applications.
The kompendium frames the alternative — secrecy — as epistemically self-defeating in this domain. EA techniques propagate through observation and imitation regardless of publication; what publication adds is the defensive vocabulary that allows targets and clinicians to recognize and resist what would otherwise propagate undetected. The asymmetry favors graduated openness: recognition-level material as widely available as the platform's general epistemic literacy work; novelty-level material restricted with audit; offensive operational material tightly controlled and accountable.
Czech-language treatment of the publication-ethics framework: /kompendium/06-etika/dual-use/. See kompendium EA for the broader framing of audience and use.
See Also
- Kompendium EA — The kompendium itself, designed under this framework
- Epistemická anihilace — Threat model whose study raises the dual-use problem
- Taxonomie EA — Tiered taxonomy structure
- Color Teams — Operational dual-use environment
- Delegitimizace — Recognition-level vocabulary
- Metakognitivní destrukce — Tightly-controlled technique class
- Kompendium — Full Czech-language reference
Connect & Contribute
Created by Tomáš Korcak (korczis) | Open Source under GHL