All articles
A focused publication on prompt injection. Direct and indirect techniques, model-specific behaviors, taxonomy, PoCs against open and closed models, defenses and their failure modes — written for working AI red teamers, not press releases.
-
Prompt Injection in AI Agents Explained
Agents turn prompt injection into tool calls, code execution and data theft. How the attack works when a model can act, and what contains it.
-
Indirect Prompt Injection: How These Attacks Work
Indirect prompt injection hides instructions in content the model retrieves. The delivery channels, the attacker objectives, and what benchmarks measure.
-
Prompt Injection Testing: A Repeatable Method
Prompt injection testing needs a method, not just a scanner. How to scope surfaces, pick a corpus and harness, define detectors, and score attack success.
-
Prompt Injection Detection Techniques That Actually Work
Compare prompt injection detection techniques: pattern filters, fine-tuned classifiers, and known-answer probes, plus why no single layer closes the hole.
-
How to Mitigate Prompt Injection: A Layered Defense Guide
No single control stops prompt injection. This guide covers the layered stack that works: input and output filtering, privilege limits, and isolation.
-
OWASP LLM Top 10 Prompt Injection (LLM01:2025) Explained
LLM01 in the OWASP LLM Top 10 is prompt injection — and it held the top spot in both the 2023 and 2025 editions. What the 2025 entry actually says.
-
Invisible Prompt Injection via Unicode Tag Smuggling
Unicode Tag characters let attackers embed invisible prompt injection payloads that still tokenize as instructions. How it works and what stops it.
-
Bing Chat Prompt Injection: The Sydney Incident
This incident analysis covers Bing Chat system-prompt extraction, persona manipulation, and indirect prompt injection through retrieved web content.
-
Garak vs PyRIT vs promptmap: Prompt Injection Testing
Garak, PyRIT, and promptmap are not interchangeable. What each prompt injection testing framework is built for, what it misses, and how to choose one.
-
Rebuff Prompt Injection: Architecture, Limits, and Gaps
Rebuff uses four prompt injection checks: heuristics, LLM classification, vector similarity, and canary tokens. This review maps their blind spots.
-
Llama 3 RAG Prompt Injection: Payloads and Defenses
A documentation-based walkthrough of four indirect prompt injection payloads for a Llama 3.1 RAG architecture and the published defenses that limit them.
-
How Prompt Injection Attacks Work: A Working Taxonomy
Prompt injection covers five classes: direct, indirect, multi-modal, agentic, and multi-turn, each with its own trust boundary and its own working defense.
-
Prompt Injection as Regulatory Failure: Deployer Liability
This analysis maps EU AI Act and GDPR duties for providers and deployers facing prompt injection, including monitoring, reporting, and documentation.