diff --git a/public/architecture.html b/public/architecture.html index 4db66657..09e1047d 100644 --- a/public/architecture.html +++ b/public/architecture.html @@ -36,9 +36,9 @@ @@ -83,7 +83,7 @@
-

Why External Enforcement May Help

+

@@ -152,7 +152,7 @@
-

The Core Hypothesis

+

Jailbreaks often work by manipulating the AI's internal reasoning. Tractatus boundaries operate external to that reasoning—the AI doesn't directly evaluate governance rules. While not foolproof, this architectural separation makes manipulation significantly harder.

@@ -163,7 +163,7 @@
-

Runtime-Agnostic Architecture

+

Tractatus works with any agentic AI system—Claude Code, LangChain, AutoGPT, CrewAI, or custom agents. The governance layer sits between your agent and its actions.

@@ -180,19 +180,19 @@
-

1. Agent Runtime Layer

+

Your AI agent (any platform). Handles planning, reasoning, tool use. Tractatus is agnostic to implementation.

-

2. Governance Layer

+

Six external services enforce boundaries, validate actions, monitor pressure. Architecturally more difficult for AI to bypass.

-

3. Persistent Storage

+

Immutable audit logs, governance rules, instruction history. Independent of AI runtime—can't be altered by prompts.

@@ -203,7 +203,7 @@
-

Six Governance Services

+

@@ -214,7 +214,7 @@
-

Boundary­Enforcer

+

Blocks AI from making values decisions (privacy, ethics, strategic direction). Requires human approval. @@ -231,7 +231,7 @@

-

Instruction­Persistence­Classifier

+

Stores instructions externally with persistence levels (HIGH/MEDIUM/LOW). Aims to reduce directive fade. @@ -248,7 +248,7 @@

-

Cross­Reference­Validator

+

Validates AI actions against instruction history. Aims to prevent pattern bias overriding explicit directives. @@ -265,7 +265,7 @@ -

Context­Pressure­Monitor

+

Monitors AI performance degradation. Escalates when context pressure threatens quality. @@ -282,7 +282,7 @@ -

Metacognitive­Verifier

+

Requires AI to pause and verify complex operations before execution. Structural safety check. @@ -299,7 +299,7 @@ -

Pluralistic­Deliberation­Orchestrator

+

Facilitates multi-stakeholder deliberation for values conflicts. AI provides facilitation, not authority. @@ -316,7 +316,7 @@

-

Explore the Architecture Interactively

+

Click any service node or the central core to see detailed information about how governance works.

@@ -353,7 +353,7 @@ -

Explore the Governance Services

+

Click any service node in the diagram (colored circles) or the central "T" to learn more about how Tractatus enforces AI safety.

@@ -365,7 +365,7 @@
-

Framework in Action

+

Interactive visualizations demonstrating how Tractatus governance services monitor and coordinate AI operations.

@@ -387,7 +387,7 @@
-

Production Reference Implementation

+

Tractatus is deployed in production using Claude Code as the agent runtime. This demonstrates the framework's real-world viability.

@@ -395,7 +395,7 @@
-

Claude Code + Tractatus

+

Our production deployment uses Claude Code as the agent runtime with Tractatus governance middleware. This combination provides:

@@ -446,7 +446,7 @@
-

Limitations and Reality Check

+

@@ -462,7 +462,7 @@

-

Known Limitations:

+

  • @@ -486,7 +486,7 @@
-

What We Need:

+

  • 🔬 @@ -520,7 +520,7 @@
    -

    Explore a Promising Approach to AI Safety

    +

    Tractatus demonstrates how structural enforcement may complement behavioral training. We invite researchers and practitioners to evaluate, critique, and build upon this work.

    diff --git a/public/locales/de/architecture.json b/public/locales/de/architecture.json index cb90fa60..17a4537c 100644 --- a/public/locales/de/architecture.json +++ b/public/locales/de/architecture.json @@ -44,32 +44,32 @@ "services": { "heading": "Sechs Governance-Dienste", "boundary": { - "name": "BoundaryEnforcer", + "name": "Boundary­Enforcer", "description": "Blockiert die KI daran, Werteentscheidungen zu treffen (Datenschutz, Ethik, strategische Ausrichtung). Erfordert menschliche Genehmigung.", "promise": "Frühes Versprechen: Wertegrenzen extern durchgesetzt – schwieriger durch Prompting zu manipulieren." }, "instruction": { - "name": "InstructionPersistenceClassifier", + "name": "Instruction­Persistence­Classifier", "description": "Speichert Anweisungen extern mit Persistenzstufen (HIGH/MEDIUM/LOW). Zielt darauf ab, Direktiven-Verfalls zu reduzieren.", "promise": "Frühes Versprechen: Anweisungen außerhalb der KI gespeichert – widerstandsfähiger gegen Kontextmanipulation." }, "validator": { - "name": "CrossReferenceValidator", + "name": "Cross­Reference­Validator", "description": "Validiert KI-Aktionen gegen Anweisungshistorie. Zielt darauf ab, Musterbias-Überschreibung expliziter Direktiven zu verhindern.", "promise": "Frühes Versprechen: Unabhängige Verifikation – KI-Behauptungen gegen externe Quelle geprüft." }, "pressure": { - "name": "ContextPressureMonitor", + "name": "Context­Pressure­Monitor", "description": "Überwacht KI-Leistungsverschlechterung. Eskaliert, wenn Kontextdruck die Qualität bedroht.", "promise": "Frühes Versprechen: Objektive Metriken können Manipulationsversuche frühzeitig erkennen." }, "metacognitive": { - "name": "MetacognitiveVerifier", + "name": "Metacognitive­Verifier", "description": "Erfordert, dass die KI pausiert und komplexe Operationen vor der Ausführung überprüft. Strukturelle Sicherheitsprüfung.", "promise": "Frühes Versprechen: Architektonische Gates zielen darauf ab, Verifikationsschritte durchzusetzen." }, "deliberation": { - "name": "PluralisticDeliberationOrchestrator", + "name": "Pluralistic­Deliberation­Orchestrator", "description": "Erleichtert Multi-Stakeholder-Beratung bei Wertekonflikten. KI bietet Moderation, keine Autorität.", "promise": "Frühes Versprechen: Menschliches Urteil erforderlich – architektonisch erzwungene Eskalation für Werte." } diff --git a/public/locales/en/architecture.json b/public/locales/en/architecture.json index 739dc528..a0ad0759 100644 --- a/public/locales/en/architecture.json +++ b/public/locales/en/architecture.json @@ -44,32 +44,32 @@ "services": { "heading": "Six Governance Services", "boundary": { - "name": "BoundaryEnforcer", + "name": "Boundary­Enforcer", "description": "Blocks AI from making values decisions (privacy, ethics, strategic direction). Requires human approval.", "promise": "Early Promise: Values boundaries enforced externally—harder to manipulate through prompting." }, "instruction": { - "name": "InstructionPersistenceClassifier", + "name": "Instruction­Persistence­Classifier", "description": "Stores instructions externally with persistence levels (HIGH/MEDIUM/LOW). Aims to reduce directive fade.", "promise": "Early Promise: Instructions stored outside AI—more resistant to context manipulation." }, "validator": { - "name": "CrossReferenceValidator", + "name": "Cross­Reference­Validator", "description": "Validates AI actions against instruction history. Aims to prevent pattern bias overriding explicit directives.", "promise": "Early Promise: Independent verification—AI claims checked against external source." }, "pressure": { - "name": "ContextPressureMonitor", + "name": "Context­Pressure­Monitor", "description": "Monitors AI performance degradation. Escalates when context pressure threatens quality.", "promise": "Early Promise: Objective metrics may detect manipulation attempts early." }, "metacognitive": { - "name": "MetacognitiveVerifier", + "name": "Metacognitive­Verifier", "description": "Requires AI to pause and verify complex operations before execution. Structural safety check.", "promise": "Early Promise: Architectural gates aim to enforce verification steps." }, "deliberation": { - "name": "PluralisticDeliberationOrchestrator", + "name": "Pluralistic­Deliberation­Orchestrator", "description": "Facilitates multi-stakeholder deliberation for values conflicts. AI provides facilitation, not authority.", "promise": "Early Promise: Human judgment required—architecturally enforced escalation for values." } diff --git a/public/locales/fr/architecture.json b/public/locales/fr/architecture.json index 8babc8b7..6efc957b 100644 --- a/public/locales/fr/architecture.json +++ b/public/locales/fr/architecture.json @@ -44,32 +44,32 @@ "services": { "heading": "Six Services de Gouvernance", "boundary": { - "name": "BoundaryEnforcer", + "name": "Boundary­Enforcer", "description": "Empêche l'IA de prendre des décisions de valeurs (confidentialité, éthique, direction stratégique). Nécessite l'approbation humaine.", "promise": "Promesse Précoce: Frontières de valeurs appliquées en externe – plus difficile à manipuler par prompting." }, "instruction": { - "name": "InstructionPersistenceClassifier", + "name": "Instruction­Persistence­Classifier", "description": "Stocke les instructions en externe avec des niveaux de persistance (HIGH/MEDIUM/LOW). Vise à réduire la dérive des directives.", "promise": "Promesse Précoce: Instructions stockées en dehors de l'IA – plus résistantes à la manipulation contextuelle." }, "validator": { - "name": "CrossReferenceValidator", + "name": "Cross­Reference­Validator", "description": "Valide les actions de l'IA contre l'historique des instructions. Vise à empêcher le biais de pattern d'outrepasser les directives explicites.", "promise": "Promesse Précoce: Vérification indépendante – les affirmations de l'IA sont vérifiées contre une source externe." }, "pressure": { - "name": "ContextPressureMonitor", + "name": "Context­Pressure­Monitor", "description": "Surveille la dégradation des performances de l'IA. Escalade lorsque la pression contextuelle menace la qualité.", "promise": "Promesse Précoce: Les métriques objectives peuvent détecter les tentatives de manipulation tôt." }, "metacognitive": { - "name": "MetacognitiveVerifier", + "name": "Metacognitive­Verifier", "description": "Exige que l'IA pause et vérifie les opérations complexes avant l'exécution. Vérification de sécurité structurelle.", "promise": "Promesse Précoce: Les portes architecturales visent à appliquer les étapes de vérification." }, "deliberation": { - "name": "PluralisticDeliberationOrchestrator", + "name": "Pluralistic­Deliberation­Orchestrator", "description": "Facilite la délibération multi-parties prenantes pour les conflits de valeurs. L'IA fournit la facilitation, pas l'autorité.", "promise": "Promesse Précoce: Jugement humain requis – escalade architecturalement appliquée pour les valeurs." }