About AI Hacking
A comprehensive AI/LLM security resource hub built for security professionals.
Our Mission
AI Hacking exists to close the gap in AI security education. As AI systems become deeply embedded in critical infrastructure, the attack surface grows — but the resources to secure them remain fragmented and outdated. We aim to be the definitive, up-to-date resource for anyone working at the intersection of artificial intelligence and information security.
We cover offensive and defensive techniques side by side: you cannot build effective defenses without understanding how attackers think. Every guide is grounded in real CVEs, reproducible techniques, and current tooling.
Who This Site Is For
Security Researchers
Investigate novel attack vectors in LLMs, agents, and AI infrastructure. Access CVE databases, reproducible proof-of-concepts, and emerging threat intelligence.
Developers
Build secure AI applications with defensive patterns for input validation, output filtering, prompt hardening, and API security.
Pentesters
Add AI systems to your assessment scope. Use our methodologies, checklists, and tooling guides to evaluate LLM deployments effectively.
Compliance Teams
Understand AI-specific regulatory requirements, from the EU AI Act to NIST AI RMF, and map them to actionable security controls.
What We Cover
Prompt Injection
Direct, indirect, and multi-turn prompt injection. Jailbreaks, encoding attacks, system prompt extraction, and defensive countermeasures.
Read guide →RAG Attacks
Document poisoning, retrieval manipulation, embedding attacks, and vector database security in Retrieval-Augmented Generation systems.
Read guide →MCP Security
Model Context Protocol vulnerabilities, server hardening, tool poisoning, and the 60+ CVEs discovered in 2026.
Read guide →Red Teaming
Structured methodologies for adversarial AI testing, from reconnaissance to reporting. Tooling and case studies included.
Read guide →Agentic AI Threats
Goal hijacking, tool misuse, supply chain attacks, and securing autonomous agents with planning and execution capabilities.
Read guide →LLM API Security
Authentication, rate limiting, key management, cost controls, and output validation for LLM-backed APIs.
Read guide →Content Freshness Policy
Content is refreshed roughly monthly, and every page carries the date it was last reviewed so you can judge how current it is. The CVE database is a curated snapshot pulled from the NVD rather than a live feed, and it shows the date it was last refreshed — for anything published in the last few days, check the NVD directly. Incident entries are written up as disclosures land, so a gap of a week or two between an incident happening and appearing here is normal.
- Weekly: CVE roundups, new tool additions, incident reports
- Monthly: Guide updates, methodology revisions, threat landscape summaries
- Quarterly: Major guide overhauls, new feature pages, structural improvements
Team
Jan Hazenberg
Maintainer & Founder
Senior Designer Voice at Odido Netherlands. Jan built AI Hacking to address the lack of centralized, practitioner-focused AI security resources. He maintains the site, curates content, and ensures accuracy across all guides.
Why This Site Exists
AI security knowledge is scattered across research papers, conference talks, and vendor blogs. AI Hacking brings it together in one place — structured, searchable, and kept current — so you spend less time searching and more time securing.