AI application security guides
Start with the business decision your team is about to make. Connecting an assistant to internal documents raises different questions from letting it approve an invoice. These guides cover both and show what evidence to keep.
Where to start
Planning a review
Use the worksheet to record the owner, expected behavior, and acceptance criteria for each feature in scope.
AI Security Assessment Checklist: 12 Tests for Agents and Connected ToolsConnecting internal documents
Follow a permission change from the source document to the assistant’s answer. Decide who owns each step.
RAG Access Control: What Happens When Document Permissions Change?Approving an action
Compare the action a person approved with the action the tool is permitted to take.
AI Tool Approval Security: Does the Executed Action Match What You Approved?More ai application security guides
- A Model Download Is Part of the Software Supply Chain
- AI Agent Permissions: Defining What a Business Workflow May Change
- AI Agent Sandbox Security: What Your Assessment Should Verify
- AI Assistant Memory Security: What Can Carry Over Into the Next Customer’s Task?
- AI Output Validation: From a Generated Answer to a Business Action
- AI Prompt Logs: Where Business Data Goes After the Answer
- AI Security Evaluations Need Different Questions From Accuracy Tests
- AI Usage Limits Are Also Availability Controls
- AI-Assisted Security Assessments: How They Work and What the Report Shows
- Knowledge Base Poisoning: Checking the Sources Behind an AI Answer
- MCP Security: Why a Valid Token May Still Be the Wrong Token
- OpenAI vs Open Source: Hacking of HuggingFace
- The AI Arms Race
- When Business Documents Become Instructions for an AI Assistant
Guides by Mario Luckeneder, founder of Ceron.