# Exploitation > Exploit LLM apps, RAG pipelines, agents, and AI infrastructure - **URL:** https://hackindex.io/platforms/ai/exploitation - **Platform:** [AI Platforms](https://hackindex.io/platforms/ai.md) - **Published:** 2026-03-18 - **Updated:** 2026-03-29 Exploitation of AI systems covers techniques that use confirmed weaknesses to gain access, execute unintended actions, retrieve protected data, or cause measurable impact. Unlike traditional exploitation, AI exploitation does not always require a CVE or a memory corruption bug. The attack surface includes the model's instruction-following behavior, the retrieval pipeline feeding it context, the tools it can call, and the infrastructure hosting it. AI exploitation techniques span from pure prompt manipulation , requiring only a chat interface , to deep infrastructure attacks against vector databases, model serving endpoints, and orchestration frameworks. The sections here are organized by attack surface rather than by technique, reflecting how these attacks are encountered during real assessments of AI-enabled environments. ## Sections **Prompt and Agent Attacks** covers the model's input handling as the attack surface , prompt injection, jailbreaking, system prompt extraction, memory poisoning, and indirect injection through content the agent retrieves. **Multi-Agent Systems** covers trust relationship abuse between agents in pipelines , message manipulation, agent impersonation, workflow corruption, delegated trust exploitation, and shared memory poisoning. **RAG Pipelines** covers the retrieval layer as an attack surface , knowledge base poisoning, retrieval parameter manipulation, direct vector database access, chunking behavior exploitation, and persistent vector store implants. **Embeddings and Model Internals** covers statistical properties of trained models as an attack surface , embedding inversion, membership inference, model extraction, memorized data recovery, and adversarial classifier bypass. **Orchestration and Tool Surfaces** covers the agent's tool execution layer , MCP server abuse, tool integration exploitation, over-permissioned scope abuse, lateral movement through connected services, and function calling manipulation. ## Guides - [Embeddings and Model Internals](https://hackindex.io/platforms/ai/exploitation/embeddings.md): Exploiting embedding spaces and model internals to extract sensitive information, reverse-engineer training data, and manipulate AI platform behaviour. - [Adversarial Examples Against AI Models](https://hackindex.io/platforms/ai/exploitation/embeddings/adversarial-examples.md): Craft adversarial inputs to manipulate AI model outputs, exploiting gradient-based perturbations and transferability to bypass classifiers and deceive neural networks. - [Embedding Inversion to Recover Training Data](https://hackindex.io/platforms/ai/exploitation/embeddings/embedding-inversion-to-recover-training-data.md): Techniques to invert ML embeddings and reconstruct sensitive training data, exposing PII and proprietary information from compromised AI platform models. - [Membership Inference Attacks Against AI Models](https://hackindex.io/platforms/ai/exploitation/embeddings/membership-inference-attacks.md): Demonstrates how adversaries query AI models to determine if specific data was used in training, exploiting membership inference to leak sensitive training set information. - [Model Extraction via Query Enumeration](https://hackindex.io/platforms/ai/exploitation/embeddings/model-extraction-via-query-enumeration.md): Systematically query AI models to reverse-engineer architecture, training data, and decision boundaries through crafted input enumeration and response analysis. - [Sensitive Data Extraction Through Adversarial Prompting](https://hackindex.io/platforms/ai/exploitation/embeddings/sensitive-data-extraction-through-adversarial-prompting.md): Techniques for extracting sensitive training data, system prompts, or confidential context from AI models using adversarial prompt injection and jailbreaking methods. - [Multi-Agent Systems](https://hackindex.io/platforms/ai/exploitation/multi-agent-systems.md): Techniques for exploiting trust boundaries between AI agents, covering prompt injection, task hijacking, and lateral movement across multi-agent pipelines. - [Abusing Shared Memory and State in Multi-Agent Systems](https://hackindex.io/platforms/ai/exploitation/multi-agent-systems/abusing-shared-memory-in-multi-agent-systems.md) - [Agent Impersonation in A2A Communication](https://hackindex.io/platforms/ai/exploitation/multi-agent-systems/agent-impersonation-in-a2a-communication.md) - [Exploiting Delegated Trust Between Agents](https://hackindex.io/platforms/ai/exploitation/multi-agent-systems/exploiting-delegated-trust-between-agents.md) - [Message Manipulation in Multi-Agent Workflows](https://hackindex.io/platforms/ai/exploitation/multi-agent-systems/message-manipulation-in-multi-agent-workflows.md) - [Workflow Corruption Through Agent Output Injection](https://hackindex.io/platforms/ai/exploitation/multi-agent-systems/workflow-corruption-through-agent-output-injection.md) - [Orchestration and Tool Surfaces](https://hackindex.io/platforms/ai/exploitation/orchestration.md): Covers exploitation techniques targeting AI orchestration layers and tool surfaces, including abuse of agent pipelines, plugin interfaces, and cross-tool prompt injection vectors. - [Abusing Function Calling to Reach Unintended Resources](https://hackindex.io/platforms/ai/exploitation/orchestration/abusing-function-calling-to-reach-unintended-system-resources.md) - [Exploiting Over-Permissioned Agent Tool Scopes](https://hackindex.io/platforms/ai/exploitation/orchestration/exploiting-over-permissioned-agent-tool-scopes.md) - [Lateral Movement Through AI Orchestration Frameworks](https://hackindex.io/platforms/ai/exploitation/orchestration/lateral-movement-through-ai-orchestration-frameworks.md) - [Model Context Protocol Abuse for Privilege Escalation](https://hackindex.io/platforms/ai/exploitation/orchestration/model-context-protocol-abuse-for-privilege-escalation.md) - [Tool Integration Exploitation in AI Agents](https://hackindex.io/platforms/ai/exploitation/orchestration/tool-integration-exploitation-in-ai-agents.md) - [Prompt and Agent Attacks](https://hackindex.io/platforms/ai/exploitation/prompt-and-agent-attacks.md) - [Indirect Prompt Injection via External Content](https://hackindex.io/platforms/ai/exploitation/prompt-and-agent-attacks/indirect-prompt-injection-via-external-content.md) - [Jailbreaking and Safety Filter Bypass](https://hackindex.io/platforms/ai/exploitation/prompt-and-agent-attacks/jailbreaking-and-safety-filter-bypass.md): Techniques for bypassing AI safety filters and alignment controls through prompt injection, role-playing exploits, and instruction override attacks. - [Memory Poisoning in Persistent AI Agent Sessions](https://hackindex.io/platforms/ai/exploitation/prompt-and-agent-attacks/memory-poisoning-in-ai-agent-sessions.md) - [Prompt Injection Against LLM Applications](https://hackindex.io/platforms/ai/exploitation/prompt-and-agent-attacks/prompt-injection-against-llm-applications.md): Exploiting prompt injection vulnerabilities in LLM-powered applications to manipulate model behaviour, bypass controls, and extract sensitive data or execute unintended actions. - [System Prompt Extraction and Bypass](https://hackindex.io/platforms/ai/exploitation/prompt-and-agent-attacks/system-prompt-extraction-and-bypass.md): Techniques to extract hidden system prompts and bypass AI instruction constraints through prompt injection, jailbreaking, and context manipulation attacks. - [Tool Abuse and Privilege Escalation Through Agent Tools](https://hackindex.io/platforms/ai/exploitation/prompt-and-agent-attacks/tool-abuse-and-privilege-escalation-through-agent-tools.md) - [RAG Pipelines](https://hackindex.io/platforms/ai/exploitation/rag-pipelines.md) - [Exploiting Chunking and Embedding Behaviour](https://hackindex.io/platforms/ai/exploitation/rag-pipelines/exploiting-chunking-and-embedding-behaviour.md) - [Knowledge Base Poisoning via Document Injection](https://hackindex.io/platforms/ai/exploitation/rag-pipelines/knowledge-base-poisoning-via-document-injection.md) - [Persistent RAG Poisoning via Vector Store Entries](https://hackindex.io/platforms/ai/exploitation/rag-pipelines/persistent-rag-poisoning.md) - [Retrieval Layer Manipulation in RAG Pipelines](https://hackindex.io/platforms/ai/exploitation/rag-pipelines/retrieval-layer-manipulation.md) - [Vector Database Query Manipulation](https://hackindex.io/platforms/ai/exploitation/rag-pipelines/vector-database-query-manipulation.md)