Framework Home | Cross-Pillar Governance | AISM | NEXUS | Challenge Lab | Dashboard
The Technology Contribution Profile separates conceptual/design evidence, implementation, validation, and enforcement. Use the Technology Card for claim-level provenance, assurance, coverage, and limits. The nine-example review and adoption backlog captures the supplied recommendations with primary source pointers, corrected control mappings, pending verification, and acceptance work. Its candidates are not certified technologies or scored organizations.
The research library records the evidence and reasoning that informed AI SAFE² over time. Individual notes retain their original publication dates, terminology, findings, and historical framework references so the evidence trail remains inspectable.
Current normative context is AI SAFE² v3.1. Historical research does not override the current framework, CP.5 profile, MCP specification binding, or current conformance requirements.
For MCP implementation decisions, use the CP.5.MCP v3.1 profile and MCP 2026-07-28 semantics.
| Note | Research article | Current-use guidance |
|---|---|---|
| 001 | Research Note: RAG Poisoning & Context Integrity | Original publication context preserved; interpret current implementation guidance through v3.1 |
| 002 | Research Note: Non-Human Identity (NHI) Sprawl & Governance | Original publication context preserved; interpret current implementation guidance through v3.1 |
| 003 | Research Note: Swarm Consensus & Cascading Failure | Original publication context preserved; interpret current implementation guidance through v3.1 |
| 004 | Research Note: Model Serialization & Supply Chain Integrity | Original publication context preserved; interpret current implementation guidance through v3.1 |
| 005 | Research Note: Indirect Prompt Injection via Memory (MINJA) | Original publication context preserved; interpret current implementation guidance through v3.1 |
| 006 | Research Note: Runtime Isolation & The "Glass Box" | Original publication context preserved; interpret current implementation guidance through v3.1 |
| 007 | Research Note: Just-in-Time (JIT) Privilege for Agents | Original publication context preserved; interpret current implementation guidance through v3.1 |
| 008 | Comparative Analysis: AI SAFE² vs. Global Standards | Original publication context preserved; interpret current implementation guidance through v3.1 |
| 009 | Research Note: Risk Assessment - Web Grounding & RAG Integrity | Original publication context preserved; interpret current implementation guidance through v3.1 |
| 010 | Research Note: Governing the 10 Core Types of AI Agents | Original publication context preserved; interpret current implementation guidance through v3.1 |
| 011 | Research Note: The Kill Switch - From Policy to "The Red Button" | Original publication context preserved; interpret current implementation guidance through v3.1 |
| 012 | Research Note: The Engineered Liability Stack: Why EFA and AI SAFE² Must Be Combined | Original publication context preserved; interpret current implementation guidance through v3.1 |
| 013 | Research Note: The 7-Layer Reference Architecture | Original publication context preserved; interpret current implementation guidance through v3.1 |
| 014 | Standard Engineering Controls for AI SAFE² Compliance | Original publication context preserved; interpret current implementation guidance through v3.1 |
| 015 | Research Note: The HEAR Doctrine -- Human Ethical Agent of Record | Original publication context preserved; interpret current implementation guidance through v3.1 |
| 016 | Research Note: Agent Replication Governance and the Identity-Multiplying Threat | Original publication context preserved; interpret current implementation guidance through v3.1 |
| 017 | Research Note 6: NHI and Orchestration - The Board and Regulator View | Original publication context preserved; interpret current implementation guidance through v3.1 |
| 018 | Research Note 5: Agentic Bias as Operational and Security Risk | Original publication context preserved; interpret current implementation guidance through v3.1 |
| 019 | Research Note 4: Protocol-Layer Supply Chain Risks in Agentic AI | Original publication context preserved; interpret current implementation guidance through v3.1 |
| 020 | Research Note 3: Agentic Ransomware and Malicious Operators | Original publication context preserved; interpret current implementation guidance through v3.1 |
| 021 | Research Note 2: Temporal Attack Profiles in Agentic AI Systems | Original publication context preserved; interpret current implementation guidance through v3.1 |
| 022 | Research Note 1: Cognitive Surface and Memory Drift in Agentic AI Systems | Original publication context preserved; interpret current implementation guidance through v3.1 |
| 023 | Research Note 023: MCP Server Security Profile | Original publication context preserved; interpret current implementation guidance through v3.1 |
| 024 | Research Note 024: MCP Consumer Protection | Original publication context preserved; interpret current implementation guidance through v3.1 |
Framework Home | Cross-Pillar Governance | AISM | NEXUS | Examples | Challenge Lab
AI SAFE² v3.1 | Cyber Strategy Institute