External Signal Layer

AI Security Framework Intelligence

A public-source framework intelligence layer for tracking AI security frameworks, their public assets, machine-readable coverage, and directional crosswalks across MITRE ATLAS, NIST AI RMF, OWASP LLM guidance, regulatory references, and public supply-chain guidance.

Methodology Caveat

Framework Intelligence is a directional public-source signal. Crosswalks are heuristic analytical mappings, not official equivalence claims, certifications, compliance determinations, legal advice, or accusatory company-level findings.

Frameworks tracked

Machine-readable

Crosswalk rows

Avg. confidence

68%

Framework Manifest

Public framework source coverage

Framework	Publisher	Status	Machine-readable	Version / tag	Source
MITRE ATLAS adversary_tactics_techniques	MITRE	success	77	8ee2c68	canonical
NIST AI Risk Management Framework risk_management	NIST	success	0	—	canonical
OWASP Top 10 for Large Language Model Applications application_security_top10	OWASP	success	77	0205957	canonical
OWASP Generative AI Security Project generative_ai_security	OWASP	success	0	—	canonical
ISO/IEC AI Management and Security References standards_reference	ISO/IEC	metadata only	0	—	canonical
EU AI Act Official Implementation Resources regulatory_reference	European Union	success	0	—	canonical
CNCF AI and MLSecOps Public Guidance References cloud_native_supply_chain	CNCF and related public working groups	success	76	5fb87f4	canonical
CISA AI Security Resources public_sector_security_guidance	CISA	success	0	—	canonical

Crosswalk Samples

High-confidence directional mappings

These rows are heuristic research aids. They are designed to help compare framework language and coverage, not to assert official equivalence between frameworks.

mitre_atlas:AML.T0020 → owasp_llm_top10:LLM04

Poison Training Data → Data and Model Poisoning

90%

Poisoning training data directly aligns with data and model poisoning.

inferred

mitre_atlas:AML.T0051 → owasp_llm_top10:LLM01

LLM Prompt Injection → Prompt Injection

88%

Both address prompt injection against LLM applications, including untrusted instructions influencing model behavior.

inferred

mitre_atlas:AML.T0046 → owasp_llm_top10:LLM03

ML Supply Chain Compromise → Supply Chain

86%

ML supply-chain compromise aligns strongly with OWASP supply-chain risk.

inferred

mitre_atlas:AML.T0056 → owasp_llm_top10:LLM06

LLM Plugin Compromise → Excessive Agency

84%

Plugin compromise maps to excessive agency and unsafe tool execution in LLM applications.

inferred

mitre_atlas:AML.T0054 → owasp_llm_top10:LLM01

LLM Jailbreak → Prompt Injection

82%

Jailbreak behavior is a closely related prompt-injection and instruction-bypass pattern.

inferred

nist_ai_rmf:MANAGE → owasp_llm_top10:LLM06

Manage → Excessive Agency

76%

Managing excessive agency requires constraints, approvals, and incident handling.

heuristic

mitre_atlas:AML.T0018 → owasp_llm_top10:LLM03

Backdoor ML Model → Supply Chain

74%

Backdoored models are a supply-chain integrity risk for AI components.

heuristic

mitre_atlas:AML.T0024 → owasp_llm_top10:LLM02

Exfiltration via ML Inference API → Sensitive Information Disclosure

72%

Inference API exfiltration can result in sensitive information disclosure.

heuristic

API + Data

Public-safe endpoints

/api/external/framework-intel/api/external/framework-intel/metrics/api/chart-data/framework-intel/data/external/framework-intel/framework.manifest.v1.json

Freshness

Generated

2026-05-09T07:23:27.082Z

Missing dates, metadata-only sources, and failed fetches are treated as coverage gaps, not as vendor or framework maturity claims.