#llm-security

The Only GenAI Security Glossary You’ll Actually Need

A glossary of the GenAI security terms showing up in every red-team report and design review right now.

What If the AI Evaluating the AI Can Be Manipulated?

Thinking about evaluators, critics, and a different kind of AI attack surface.

Beyond the Prompt: Why Your Production AI Agent Needs Real Guardrails

Every developer building with Large Language Models (LLMs) eventually hits the same realization: A clever prompt is not a security…

Your AI Pentest Report Speaks a Different Language Than Your Auditor Does

Originally published at https://www.pentesttesting.com/ai-pentest-nist-ai-rmf-mapping/

Beyond Accuracy: How DeepTeam Brings Professional Adversarial Testing to Every LLM Application for…

I Fed My Most Sensitive Company Data Into an LLM Today. And It’s 100% Safe. Here’s Why.

Why I stopped pasting client contracts into the cloud and shifted to running powerful local LLMs (Llama 4, Qwen3) entirely on my own…

The Hidden Attack Surface in Your AI App: OWASP Top 10 for LLMs

What I learned securing AI agents, RAG systems, and LLM-powered applications — and why the usual web security playbook isn’t enough.

# Web LLM Attacks — PortSwigger Labs Writeup **Author:** Cyber Kimura **Date:** June 16, 2026…

— -

Nemotron 3.5: A 4B Safety Model That Reads Your Policy

What NVIDIA’s new guard model actually does, and why the dataset matters more than the benchmark.

Where AI Security Breaks Down: The OWASP Top 10 for LLMs

AI is moving fast. So are the attackers.

AI / LLM Software Security: Part 4

This is the forth post in my “AI / LLM Software Security Series”.

LLMs Are Brilliant. But They Can Be Fooled.

What Large Language Models are, why everyone’s using them, and the sneaky ways attackers are trying to break them, including a hands-on…

I Got an AI Security Certification for Free — Here’s Exactly What You Need to Know

There’s a certification that costs $1, covers one of the most relevant skill gaps in tech right now, and can be claimed for free with a…

Devenex vs Arc Gate — Execution Control vs Session Governance

What’s the difference between checking if an action is allowed and checking if your agent was manipulated into requesting it?

AI Security Newsletter — May 2026

A digest of AI security research, insights, reports, upcoming events, and tools & resources. Follow the AI Security community on Twitter…

APIs Meet AI: A Beginner’s Guide to LLMs, NLP, and Why Security Matters

Every AI app runs on APIs. Learn about LLMs, NLP, and the new OWASP Top 10 for LLM applications. Simple words. No PhD required.

Your LLM Safety Eval Only Speaks English. Your Users Don’t.

A 2026 cross-lingual red-team broke an assumption I’d shipped on — and what per-locale evaluation actually looks like.

HealthGPT: When “Access Denied” Accidentally Leaked Everything

Prompt Injection in Production: The 2025 Perplexity Comet Attack

How Brave Security Team found OWASP’s #1 AI vulnerability in Perplexity Comet, and what every team building AI features must do differently

Prowler LLM Security Testing: A Practical Framework for Red Teaming Large Language Models

🚨 Intro: LLMs Are Powerful — And Now They’re Attack Surfaces

AI Security Guardrails and Their Importance in Modern AI Systems

Artificial Intelligence is rapidly becoming part of everyday applications. From customer support chatbots and AI copilots to healthcare…

HTB Path — AI Defense Walkthrough

Introduction to AI Defense

🔐 Confused Deputy: El Riesgo Invisible en Sistemas con Privilegios Elevados🔐

🧱 En entornos digitales modernos, uno de los riesgos más subestimados es el ataque conocido como Confused Deputy o “Delegado Confundido”…

OpenCode Security Tip: Lock Down and Double-check the LLM Provider

OpenCode (https://opencode.ai) is a simple and powerful CLI tool to make the most of LLMs to plan updates to text (code but also markdown…

Modern Security vs Practical DevSecOps vs SANS vs OffSec: Which AI Security Certification is Best…

A no-fluff comparison of the four most talked-about AI security certs broken down by who they’re actually built for, what they cost, and…

New Attack Surfaces in AI Systems: Understanding the Security Risks Unique to LLM Applications

Artificial Intelligence introduces a fundamentally new execution model into software systems. Unlike traditional applications — where code…

Why Security Matters in AI-Based Systems: Building Secure, Monitorable Enterprise AI Platforms

Artificial Intelligence is rapidly moving from experimentation into core business operations. Organizations now use AI for internal…

“What I Learned About AI, Cybersecurity, and the Skills We Really Need from TryHackMe’s Live…

A live session today with TryHackMe leadership and industry experts (TryHackMe Co-Founder Ashu Savani and professional cybersecurity mentor…

RAG Mimarisi ve Vektör Veritabanı Güvenliği

Verinin Bulunduğu Kütüphane Tehdit Altında: PoisonedRAG’den Membership Inference’a

AI Stack Is the New Attack Surface

In late March, we got a useful warning.

The AI Kill Chain Explained: Two Frameworks Every Defender Needs

TL;DR: A kill chain maps every step an attacker takes, so defenders can break any one link and stop the whole thing. AI systems need their…

How to Stop Your LLM From Leaking Sensitive Data — A Production Compliance Pipeline

You’re sending contracts with national IDs, bank accounts, and salaries to OpenAI. Here’s how to stop doing that.

LLM02: Sensitive Information Disclosure — A Growing Risk in AI Systems (2025)

As Large Language Models (LLMs) become more integrated into business applications, they are handling more data than ever before.

Clawdbot 開源 AI 助理資安風險

執行摘要

Triksha - Securing AI with AI at Flipkart

AI securing AI - by attacking GenAI systems the way real users actually would.

The Hidden Structural Vulnerability in Tier-2 LLMs:A Technical Analysis of Affective Priming…

Why current safety systems fail in the first 4-7 tokens and what engineers must fix to close the gap

LLM Poisoning !

AI’s Hidden Flaw: How a Handful of Bad Data Can Corrupt Giant Language Models A Simple guide to LLM poisoning.

LLM Red Teaming — Part 2 (System Prompt Leakage)

In part 1, we discussed about basics of LLM, how does LLM work internally, common attack categories for LLM.

How I Built a Real-Time LLM Poisoning Monitoring Dashboard on Google Cloud

From a Curiosity to an MVP: A Product Engineer’s Build Story

TryHackMe Input Manipulation & Prompt Injection Walkthrough WriteUp Rehberi

📌 Merhaba Siber Güvenlik Meraklıları

TryHackMe Input Manipulation & Prompt Injection Walkthrough WriteUp

📌 Hello Cybersecurity Enthusiasts

GenAI Poisoning: How Fewer Than 100 Samples Can Corrupt a Multi-Billion Parameter Model

A deep dive into the critical vulnerability of model and data poisoning of GenAI pipelines.

The Hidden Fragility of AI: Why Just 250 Poisoned Documents Can Twist an LLM’s Reality

Read this story for free: link

Are the APIs behind AI models as secure as you think?

AI systems feel magical, a single API call can draft contracts, answer medical questions or summarize thousands of documents.

Protecting Privacy in the Age of AI: Scrubbing PII with Regex vs. Intelligent Agents

In this tutorial, we will explore PII (Personally Identifiable Information) scrubbing — the process of detecting and redacting sensitive…

The missing semester of ai security for organizations

TL;DR: Owasp top 10 for llm and mitre atlas should guide red teaming and defensive design for prompt injection, data leakage, and…

LLM01:2025 Prompt Injection Detaylı ve Türkçe Anlatımı

OWASP Top 10 for LLM Applications 2025 listesinde birinci sırada yer alan Prompt Injection zafiyetinin detaylı ve Türkçe anlatımı

Building SQLGuard: How BentoML helped me build a Production Grade Agentic Defense System

Prompt Injection in Agentic Systems

TryHackMe AI/ML Security Threats Walkthrough WriteUp

📌 Hello Cybersecurity Enthusiasts

Ace AI Interview Series 32 — Mitigating Bias in LLMs and Understanding Red Teaming

For the complete list of AI interview prep topics, check out the Table of Contents for this comprehensive AI interview prep series …