aisecurity.llc
Research Channels
Eight independent monitoring channels. One taxonomy. Zero overlap. When academic research, open-source activity, press coverage, threat disclosures, and adversary frameworks all point at the same gap — that convergence is the argument.
Monitoring channels
Independent research channels
The report triangulates across the job description corpus (what employers say they need), practitioner surveys (what practitioners are experiencing), and the monitoring channels below. Each channel is independently collected and classified against the same AI security taxonomy. When multiple independent sources describe the same structural gap from different angles, the convergence is the argument — not any single dataset alone.
MITRE ATLAS Navigator
Tactics, techniques, mitigations, and case studies curated from the upstream atlas-data repository. Release v5.6.1 (5.6.0).
Academic Velocity
Paper velocity, bucket-share composition, and term acceleration across AI security research domains. What academia studies today is practitioner vocabulary in 12–24 months.
Open Source Velocity
Repo growth, contributor density, and event activity across AI security domains. Open-source builder attention is the market's most honest leading indicator.
Press Coverage
Industry press classified across AI security taxonomy buckets. Tracks how the field is being framed to executive audiences — and where that framing diverges from operational reality.
Concept Maturity
A discipline-maturity clock: when AI security concepts acquire institutional codification, junior talent pipelines can form. This channel tracks where that transition is — and isn't — happening.
Threat Disclosure
CVE, NVD, GHSA, OSV, and CISA KEV records classified for AI/ML relevance. Tracking the gap between research discovery and formal public disclosure.
Controls Crosswalk
Directional mappings across MITRE ATLAS, NIST AI RMF, OWASP LLM Top 10, and related governance references. Framework fragmentation is itself a signal.
Convergence
Where independent channels agree. Cross-source composite scoring surfaces the highest-confidence trends across academic research, builders, press, threat disclosure, and framework intelligence.
Signal inventory
Skills, tools, and framework signals
The benchmark tracks 85+ named signals from job description corpora — skills, tools, frameworks, attack surfaces, and psychometric patterns that define the AI security labor market.
| Rank | Signal | Family | Why it matters |
|---|---|---|---|
| 1 | Prompt injection and indirect prompt injection | Attack surface | A flagship AI-specific attack surface for report and lab content. |
| 2 | RAG security and retrieval controls | Architecture | Bridges data access, context control, and application security. |
| 3 | Agent and tool-calling authorization | Agent security | Core to the Agentic Anarchy finding. |
| 4 | AI evals and adversarial testing | Evidence | Turns model behavior into repeatable security evidence. |
| 5 | NIST AI RMF and ISO 42001 mapping | Governance | Important only when connected to evidence and engineering artifacts. |