Topics
Browse everything AI Attacks publishes by category and tag: every topic covered here, with the most recent guides listed under each one.
Tags
- #llm-security 8
- #prompt-injection 8
- #adversarial-ml 7
- #red-teaming 7
- #jailbreaking 6
- #attack-patterns 4
- #supply-chain 3
- #agent-security 2
- #black-box-attacks 2
- #data-poisoning 2
- #evasion-attacks 2
- #indirect-prompt-injection 2
- #model-security 2
- #adversarial-examples 1
- #adversarial-patches 1
- #adversarial-suffix 1
- #agentic-ai 1
- #ai-attacks 1
- #ai-security 1
- #backdoor-attacks 1
- #benchmarking 1
- #ci-cd 1
- #context-poisoning 1
- #dataset-integrity 1
- #defense 1
- #evasion 1
- #fgsm 1
- #fine-tuning 1
- #fraud-detection 1
- #function-calling 1
- #garak 1
- #gcg 1
- #gpt-4v 1
- #hugging-face 1
- #indirect-injection 1
- #ip-theft 1
- #llm-attacks 1
- #long-context 1
- #machine-learning-security 1
- #malware-detection 1
- #many-shot 1
- #membership-inference 1
- #memory-attacks 1
- #model-extraction 1
- #model-inversion 1
- #model-poisoning 1
- #model-robustness 1
- #multimodal 1
- #optimization-attacks 1
- #owasp 1
- #persistence 1
- #privacy-attacks 1
- #rag 1
- #rag-poisoning 1
- #rag-security 1
- #regression-testing 1
- #retrieval-augmented-generation 1
- #safety-training 1
- #tool-call-hijacking 1
- #training-data 1
- #training-data-extraction 1
- #vector-database 1
- #vision-models 1
- #visual-prompt-injection 1
- #vulnerability-taxonomy 1
- #web-scraping 1
- #white-box 1
Categories
Adversarial ML 5 posts
- Evasion Attacks in Production ML: Malware, Spam, and FraudDeployed ML classifiers in malware, spam, and fraud detection face evasion attacks where the attacker has a clear payoff.
- Web-Scale Dataset Poisoning: Split-View and FrontrunningSplit-view attacks exploit mutable web content, while frontrunning targets timed snapshots to poison datasets without training pipeline access.
- Adversarial Examples Against Vision Models in 2026Where physical-world adversarial patches and digital attacks stand against modern vision models, what has been hardened, and what red teams should test.
- Model Extraction via Black-Box Query AttacksHow attackers reconstruct model behavior and decision boundaries through query-only access, the economics of the attack, and what stolen models enable.
- Model Inversion and Membership Inference in LLMsHow membership inference reveals whether specific data trained a model, and how inversion reconstructs private text from gradients and model outputs.
Attack Patterns 5 posts
- Indirect Prompt Injection in RAG PipelinesHow attackers embed malicious instructions in documents that get retrieved into LLM context — and why RAG makes prompt injection a supply-chain problem.
- Jailbreaking Multimodal Models: Visual Prompt InjectionHow attackers use images, typography, and adversarial perturbations to bypass safety guardrails in multimodal models, and why vision widens the surface.
- LLM Context Window PoisoningHow attackers plant persistent malicious instructions in memory and context, survive summarization, and carry influence across separate LLM sessions.
- Supply Chain Attacks on AI Models: Poisoning and BackdoorsHow attackers compromise models before production through dataset poisoning, malicious fine-tuning, and serialization exploits, and how to detect it.
- Tool-Call Hijacking in Agentic SystemsHow attackers exploit the gap between LLM reasoning and tool execution to trigger unauthorized tool calls, and why confirmation gaps are the root cause.
Attack Techniques 5 posts
- Prompt Jailbreak Examples: 7 Patterns ExplainedSeven prompt jailbreak examples from published research, explained by mechanism: what each does to the model and how defenders catch it.
- How Indirect Prompt Injection WorksHow indirect prompt injection works: attacker instructions hidden in web pages, emails and RAG documents, the attack surface, and what defenders can do.
- LLM Jailbreak Examples: 10 Documented PatternsTen LLM jailbreak examples drawn from published research, with the reported success rates, the mechanism behind each, and the signals that detect them.
- How GCG Works: Greedy Coordinate Gradient, ExplainedA mechanics-first look at Greedy Coordinate Gradient (GCG): how the algorithm optimizes tokens over discrete space, why suffixes transfer, and its limits.
- Many-Shot Jailbreaking: What the Research ShowsHow hundreds of synthetic compliance examples in a long-context prompt erode safety training, what the published results show, and why patching is hard.
LLM Security 2 posts
- RAG Poisoning: Corrupting Retrieval and EmbeddingsRAG poisoning corrupts the retrieval corpus or embedding store so a model pulls attacker-controlled context at inference, without touching its weights.
- OWASP Top 10 LLM Explained: Every Entry and What to FixThe OWASP Top 10 for LLM Applications 2025 explained: every entry, what it means in production, what the revision changed, and which fixes return most.