I'm Emmanuel Guilherme — an AI security leader who thinks like an auditor. I lead the Data Security Initiative (DSGAI) of the OWASP GenAI Security Project, co-lead LLM02 (Sensitive Information Disclosure) on the OWASP Top 10 for LLM Applications, and serve as a Candidate Expert on Canada's mirror committee for ISO/IEC JTC 1/SC 42 — the committee behind ISO/IEC 42001, 23894, and 42005. By day: Senior Global IT Auditor at Martinrea International (Tier 1 automotive).

Work I've contributed to at OWASP is cited in joint AI security guidance from the governments of the USA, Canada, the UK, Australia, and New Zealand, and referenced by CISA on agentic AI adoption.

I work where AI security meets enterprise audit reality. The repos here are open-source controls, mappings, and tooling that translate emerging GenAI risks into language CISOs, auditors, and regulators can act on.

Based in North America. Trilingual: English, Portuguese, Spanish.

• OWASP GenAI Security Project — Lead, Data Security Initiative (DSGAI01–DSGAI21): one of three principal authors, driving a global community of 240+ contributors.
• OWASP Top 10 for LLM Applications — Co-lead, LLM02 (Sensitive Information Disclosure); author of Persistent Memory Poisoning, a candidate entry for the next edition.
• ISO/IEC JTC 1/SC 42 (Canada Mirror Committee) — Candidate Expert on AI standards (42001 · 23894 · 42005).
• Threat Modeling Connect, Toronto Chapter — Co-founder.
• Packt — AI Security Technical Reviewer (LLM security · adversarial ML · AI threat modeling).

• GenAI Data Security Risks and Mitigations 2026 (DSGAI01–DSGAI21) — principal author, initiative lead. Interactive guide.
• LLM Applications Cybersecurity & Governance Checklist v1.1 — contributor. Referenced in joint government guidance on deploying AI systems securely (USA · Canada · UK · Australia · New Zealand).
• GenAI Red Teaming Guide — co-author.
• LLM & GenAI Security Center of Excellence Guide — co-author.
• LLM & GenAI Security Solutions Landscape — co-author.
• Agentic AI — Threats and Mitigations — contributor.

• Cited by CISA — Careful Adoption of Agentic AI Services.
• Featured in Lakera's GenAI Security Readiness Report (2024, 2025).
• Quoted in the AIRQ (AI Risk Quadrant) framework alongside leaders from CSA and CoSAI.

Four sibling repos covering the governance, audit, offensive, and shadow-AI sides of enterprise AI security. Standards-aligned across ISO/IEC 42001, NIST AI RMF, EU AI Act, OWASP, MITRE ATLAS, SOC 2, OSFI E-21, and NYDFS Part 500. MIT-licensed code, CC-BY 4.0 content. All browser-only — no backend, no telemetry — each installable as an offline-capable PWA.

AI-Governance-Toolkit · live app — Stand up AI governance by Friday. AI system inventory with EU AI Act risk classification, AI Impact Assessment per ISO/IEC 42005, risk register with heatmaps, vendor risk questionnaire (40 weighted questions, red-flag detection), maturity self-assessment (8 domains, radar chart, action plan), interactive ISO/IEC 42001 roadmap (52-week tracker), and 12 downloadable templates. Word/Excel export.

AI-Controls-Catalog · live app — Audit-ready controls library for AI systems. 20 controls across 14 categories with Test of Design and Test of Operating Effectiveness guidance, sample-size direction, evidence requirements, and mappings to 10 frameworks. Scoping wizard (6-step questionnaire → tailored control set), RCM Excel export (4-worksheet working paper), Word/PDF/CSV/JSON export.

AI-RedTeam-Framework · live app — Stand up an AI red team at a regulated enterprise. 15-chapter playbook (charter → maturity model), 25 attack patterns mapped to OWASP LLM/Agentic Top 10, MITRE ATLAS, and NIST AI RMF, 15 candid tool reviews, 10 downloadable templates, and an interactive Engagement Planner that generates Word Rules of Engagement.

Shadow-AI-Defense · live app — Detect and respond to Shadow AI in your enterprise. 35 cataloged AI services with network and client signatures, 30 detection rules (Sentinel KQL, Defender for Cloud Apps, Defender for Endpoint, Purview DLP, CrowdStrike Falcon, Netskope, Zscaler, Conditional Access, generic network), 8 graduated response runbooks with RACI matrices, 8 security stack profiles, communications templates, and a policy starter.

GenAI & Agentic AI Incidents — 12,500+ documented GenAI and agentic AI security incidents, each cross-referenced to OWASP LLM Top 10 (2025), Agentic Top 10, NIST AI RMF, and MITRE ATLAS. DOI-registered (10.5281/zenodo.20248676), published on PyPI and Hugging Face, with STIX 2.1, TAXII, and MISP feeds for threat-intel platforms.

GenAI-Security-Crosswalk — Open-source mapping of OWASP GenAI risks — LLM Top 10, Agentic Top 10, and DSGAI 2026 — to 25 industry frameworks, including NIST AI RMF, ISO/IEC 42001, EU AI Act, MITRE ATLAS, and OT/ICS guidance. 3,351 mappings across 1,016 controls; also published on npm.

DSGAI — Interactive web guide for the OWASP GenAI Data Security Risks and Mitigations 2026 publication I lead — 21 risk entries (DSGAI01–DSGAI21) with a three-tier mitigation model. Companion repo: GenAI-Data-Security-Initiative — taxonomies, crosswalks, datasets, and tooling for securing data in GenAI systems.

GenAI-Security-Literature-Review — Community-driven, auto-updating literature review of GenAI/LLM security research. 240+ curated entries across 46 categories, with weekly automated discovery from arXiv, Semantic Scholar, and CrossRef.

finagent-redrange — Reproducible, defensive red-team range for financial-services AI agents: nine proof-of-concept exploits against a mock banking agent — covering the full OWASP LLM Top 10 — each paired with the control that blocks it and mapped to MITRE ATLAS and NIST AI RMF.

AgentVulnMitigator — agentvuln, a runtime guardrail SDK for agentic AI. Normalizes evasive input (zero-width characters, confusables, base64, spacing tricks) and blocks prompt injection, jailbreaks, data exfiltration, and unsafe tool arguments before they reach the model or its tools.

• Data security for GenAI — DSGAI 2026 (training data governance, inference-time leakage, RAG, agentic workflows)
• LLM & agentic AI risk — prompt injection, tool misuse, autonomy scope, multi-agent threat modeling
• AI governance & audit — ISO/IEC 42001, NIST AI RMF, EU AI Act, MITRE ATLAS, audit-ready control mappings
• Shadow AI & enterprise controls — detection engineering, DLP, OT/ICS implications of AI deployment

• OWASP GenAI Security Project
• LinkedIn — the fastest channel for standards collaboration, DSGAI contributions, or speaking invitations.