About HexTyx — AI Security Research Team

Our Mission

AI systems are being deployed at enterprise scale without the security controls that traditional software has had for decades. Most security teams know how to test networks and applications. Almost none have a systematic way to test the AI systems they're now responsible for. HexTyx exists to close that gap — with a platform that tests AI deployments against the same structured threat intelligence framework (MITRE ATLAS) that the security community uses to map traditional threats.

144

Research Articles

Free Security Tools

40+

ATLAS Techniques Mapped

Industries Benchmarked

102

Page Benchmark Report

2026

Founded

The HexTyx Research Teams

Our research and content is produced by four specialist teams under the HexTyx Security Research umbrella. Each team owns a distinct domain — ensuring depth over breadth on every topic we cover.

Security Research Team

Core vulnerability research, attack pattern analysis, MITRE ATLAS technique mapping, and adversarial testing methodology.

Covers: Attack vectors, red teaming, prompt injection, agent security, threat landscape

️

Engineering Team

Scan engine architecture, API security, RAG pipeline security, DevSecOps integration, and tool implementation research.

Covers: RAG security, agent architecture, SaaS security, MLOps, technical guides

Compliance & Governance Team

Regulatory framework analysis, OWASP mapping, EU AI Act compliance, enterprise risk classification, and audit methodology.

Covers: EU AI Act, NIST AI RMF, SOC 2, ISO 27001, FedRAMP, GDPR, PCI DSS

Growth & Intelligence Team

Industry benchmarking, competitive analysis, market research, and AI security landscape intelligence.

Covers: Benchmark research, tool comparisons, industry analysis, security market trends

Research Methodology

Every piece of content and every tool we publish follows a structured methodology grounded in established security frameworks — not opinions.

1

MITRE ATLAS Mapping All vulnerability research is mapped to the MITRE ATLAS framework (Adversarial Threat Landscape for AI Systems) — the authoritative taxonomy for AI/ML-specific attack techniques. We reference specific technique IDs (e.g. AML.T0051) throughout our research.
2

OWASP LLM Top 10 Alignment Our security assessments and compliance content align with the OWASP LLM Top 10 — the practitioner standard for LLM application security. Our free assessment tool scores organisations across all 10 categories.
3

Proof-of-Exploit Validation Our scanner produces Proof-of-Exploit evidence — confirmed attack success, not theoretical vulnerability presence. We test across 7 attack categories with payload mutation to catch both direct and evasion-based attacks.
4

Benchmark Rigour The HexTyx AI Security Benchmark 2026 analysed deployments across 8 industries using a consistent 23-phase methodology. Coverage scores reflect actual MITRE ATLAS technique coverage, not self-reported maturity assessments.
5

Continuous Update Commitment AI attack techniques evolve rapidly. We update research articles when new attack patterns emerge, new framework versions are published, or when real-world incidents reveal gaps in our prior analysis. Every article shows both publish and update dates.

Frameworks & Standards We Follow

MITRE ATLAS MITRE ATT&CK OWASP LLM Top 10 NIST AI RMF NIST CSF 2.0 EU AI Act ISO/IEC 27001 ISO/IEC 42001 SOC 2 Type II FedRAMP PCI DSS v4 GDPR HIPAA CCPA

The Platform

HexTyx is built on the AIZA-Cortex platform — a network of AI-native security and intelligence products under the AIZA brand. The platform is designed for the DC/NOVA enterprise and government security market, with a research architecture that supports both self-service testing and enterprise red team engagements.

The scan engine tests AI deployments against 7 attack categories using real HTTP requests to target endpoints — not static prompt analysis. It produces MITRE ATLAS-mapped findings, Proof-of-Exploit confirmation, and reports in JSON, PDF, SARIF, STIX, and MITRE ATT&CK Navigator formats.

Current status: Phase 1 — free tools, research, and benchmark are live. The full scan platform is in closed beta. Join the waitlist →