# AI SAFE² Research Index
### Evidence, threat analysis, and framework foundations

[![AI SAFE² v3.1](https://img.shields.io/badge/AI_SAFE%C2%B2-v3.1-F6921E?style=flat-square)](../README.md)
[![Surface: Research](https://img.shields.io/badge/Surface-Research-820F1A?style=flat-square)](./README.md)
[![Context: v3.1 Current](https://img.shields.io/badge/Context-v3.1_Current-808080?style=flat-square)](../docs/REPOSITORY-UX-STANDARD.md)

[Framework Home](../README.md) | [Cross-Pillar Governance](../00-cross-pillar/README.md) | [AISM](../AISM/) | [NEXUS](../NEXUS/) | [Challenge Lab](../challenges/) | [Dashboard](https://cyberstrategyinstitute.github.io/ai-safe2-framework/dashboard/)

---

## Technology contribution reviews

The [Technology Contribution Profile](../AISM/TECHNOLOGY-CONTRIBUTION-PROFILE.md)
separates conceptual/design evidence, implementation, validation, and enforcement.
Use the [Technology Card](../AISM/templates/TECHNOLOGY-CARD.md) for claim-level
provenance, assurance, coverage, and limits. The
[nine-example review and adoption backlog](./technology-contribution-examples.md)
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.

## How to read the research library

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](../00-cross-pillar/cp5_mcp_server_security.md) and MCP `2026-07-28` semantics.

---

## Research notes

| Note | Research article | Current-use guidance |
|---|---|---|
| 001 | [Research Note: RAG Poisoning & Context Integrity](./001_rag_poisoning.md) | Original publication context preserved; interpret current implementation guidance through v3.1 |
| 002 | [Research Note: Non-Human Identity (NHI) Sprawl & Governance](./002_nhi_secret_sprawl.md) | Original publication context preserved; interpret current implementation guidance through v3.1 |
| 003 | [Research Note: Swarm Consensus & Cascading Failure](./003_swarm_consensus_failure.md) | Original publication context preserved; interpret current implementation guidance through v3.1 |
| 004 | [Research Note: Model Serialization & Supply Chain Integrity](./004_supply_chain_model_signing.md) | Original publication context preserved; interpret current implementation guidance through v3.1 |
| 005 | [Research Note: Indirect Prompt Injection via Memory (MINJA)](./005_memory_injection_minja.md) | Original publication context preserved; interpret current implementation guidance through v3.1 |
| 006 | [Research Note: Runtime Isolation & The "Glass Box"](./006_runtime_isolation_gvisor.md) | Original publication context preserved; interpret current implementation guidance through v3.1 |
| 007 | [Research Note: Just-in-Time (JIT) Privilege for Agents](./007_jit_privilege_access.md) | Original publication context preserved; interpret current implementation guidance through v3.1 |
| 008 | [Comparative Analysis: AI SAFE² vs. Global Standards](./008_grc_framework_comparison.md) | Original publication context preserved; interpret current implementation guidance through v3.1 |
| 009 | [Research Note: Risk Assessment - Web Grounding & RAG Integrity](./009_web_grounding_risk.md) | Original publication context preserved; interpret current implementation guidance through v3.1 |
| 010 | [Research Note: Governing the 10 Core Types of AI Agents](./010_governing_agent_types.md) | Original publication context preserved; interpret current implementation guidance through v3.1 |
| 011 | [Research Note: The Kill Switch - From Policy to "The Red Button"](./011_the_kill_switch.md) | 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](./012_the_engineered_liability_stack.md) | Original publication context preserved; interpret current implementation guidance through v3.1 |
| 013 | [Research Note: The 7-Layer Reference Architecture](./013_the_7_layer_stack.md) | Original publication context preserved; interpret current implementation guidance through v3.1 |
| 014 | [Standard Engineering Controls for AI SAFE² Compliance](./014_compensating_controls_spec.md) | Original publication context preserved; interpret current implementation guidance through v3.1 |
| 015 | [Research Note: The HEAR Doctrine -- Human Ethical Agent of Record](./015_hear_doctrine_human_ethical_agent_of_record.md) | Original publication context preserved; interpret current implementation guidance through v3.1 |
| 016 | [Research Note: Agent Replication Governance and the Identity-Multiplying Threat](./016_agent_replication_governance.md) | Original publication context preserved; interpret current implementation guidance through v3.1 |
| 017 | [Research Note 6: NHI and Orchestration - The Board and Regulator View](./017_nhi_orchestration_board_and_regulator_view.md) | Original publication context preserved; interpret current implementation guidance through v3.1 |
| 018 | [Research Note 5: Agentic Bias as Operational and Security Risk](./018_agentic_bias_as_operational_risk.md) | Original publication context preserved; interpret current implementation guidance through v3.1 |
| 019 | [Research Note 4: Protocol-Layer Supply Chain Risks in Agentic AI](./019_protocol_layer_supply_chain_risks.md) | Original publication context preserved; interpret current implementation guidance through v3.1 |
| 020 | [Research Note 3: Agentic Ransomware and Malicious Operators](./020_agentic_ransomware_and_malicious_operators.md) | Original publication context preserved; interpret current implementation guidance through v3.1 |
| 021 | [Research Note 2: Temporal Attack Profiles in Agentic AI Systems](./021_temporal_attack_profiles.md) | Original publication context preserved; interpret current implementation guidance through v3.1 |
| 022 | [Research Note 1: Cognitive Surface and Memory Drift in Agentic AI Systems](./022_cognitive_surface_and_memory_drift.md) | Original publication context preserved; interpret current implementation guidance through v3.1 |
| 023 | [Research Note 023: MCP Server Security Profile](./023_mcp-server-security-profile.md) | Original publication context preserved; interpret current implementation guidance through v3.1 |
| 024 | [Research Note 024: MCP Consumer Protection](./024_mcp_consumer_protection.md) | Original publication context preserved; interpret current implementation guidance through v3.1 |

---

## Repository navigation

[Framework Home](../README.md) | [Cross-Pillar Governance](../00-cross-pillar/README.md) | [AISM](../AISM/) | [NEXUS](../NEXUS/) | [Examples](../examples/) | [Challenge Lab](../challenges/)

*AI SAFE² v3.1 | Cyber Strategy Institute*
