About

AnShinGPT publishes practical guides for teams that ship LLM and AI agent features to production. The guides come from the team behind AnShinGPT, a GenAI safety scanner API that checks prompts and model responses for jailbreaks, prompt leaks, personal data and unsafe content.

What you will find here

• Security fundamentals such as LLM firewalls, guardrails and gateways
• Prompt injection defences, from direct attacks to indirect and multilingual variants
• RAG, agent and MCP security, including access control and tool-call limits
• Output and data protection guidance, plus checklists for evaluation and operations

Each article opens with a short answer, explains the mechanism, and ends with frequently asked questions.

How we write our guides

1. Start with the question. Each guide answers a real question a builder would ask.
2. Keep it practical. We prefer concrete controls and examples to vague advice.
3. Be honest about limits. No single control stops every attack, so we describe layered defences.
4. Keep it current. Guides are reviewed as the threats and standards change.

A note on advice

These guides are general information, not a security audit. Test any control against your own system and threat model before relying on it.

Get in touch

Found a mistake, or want a topic covered? Visit our contact page.