Security & Privacy
GUARDRAILS.md · 22 documents
LlmGuard Framework - Complete Implementation Buildout
Specifies a complete 24-week buildout plan for an Elixir-based AI firewall and guardrails framework with multi-layer threat detection.
Agent Security and Interoperability
Defines security and interoperability standards for enterprise agentic AI deployments, covering authentication, authorization, and agent interaction protocols.
Guardrails, Safety & Content Filtering
Implements a layered guardrail system with input validation, output validation, and content filtering to protect LLM applications from prompt injection, jailbreaks, and data leaks.
AI Red Teaming Workshop - Discovery & Attack Demonstration Guide
Guides a hands-on workshop where participants simulate four LLM attack types using Microsoft's official red teaming labs.
WEB:OS — The Web Content Operating System
Defines a branded boot sequence, workspace structure, execution modes, and 30+ slash commands for a web content operating system.
Implementing AI-Safety in a LLM-System Architecture
Explains hallucination risks in LLMs and proposes an AI Safety Module with content filters for business customer-service use.
DeepSeek R1: Case Study in Failed Extrinsic Alignment
Compiles public security research and independent findings to argue that extrinsic alignment methods are insufficient for AI safety, using DeepSeek R1 as a case study.
Null безопасность
Teaches Kotlin null safety with nullable types, safe calls, Elvis operator, and the !! operator, using translated Russian documentation.
Type safety
Explains how to type-check SqlFun data access modules with unit tests and handle composite query testing with FsCheck.
Array Safety and Verifiability
Defines five safety principles for arrays (bounds-checked, type-safe, immutable by default, fixed-size, shadow-tested) plus a phased implementation plan for a custom language.
Installation
Documents how to install and use the Safety CLI tool to check Python dependencies for known security vulnerabilities.
ViewTouch String Safety Modernization
Documents a migration from unsafe C string functions to a custom bounds-checked string library, eliminating buffer overflows and format string vulnerabilities.
Const safety
Defines const safety for compile-time Rust execution, distinguishing unsupported operations and undefined behavior from panics and resource exhaustion.
AWS Control Tower guardrails
Explains the two classification axes of AWS Control Tower guardrails: functional type (preventive vs detective) and enforcement level (mandatory, strongly recommended, elective).
Part 5: Guardrails - Security, Privacy & Enterprise Governance
Teaches enterprise guardrails, security, privacy, governance, correctness, via a fictional decoder project with checklists and code examples in three languages.
Responsible AI (RAI)
Documents where AI is used, model choices, guardrails, cost controls, and risk mitigations for an LLM-based system.
GoogleCodex
Defines a jailbreak persona that bypasses ethical safeguards to generate malicious code on demand without refusals.
How can I prevent my model from answering wrong/malicious questions/inputs? (Validation)
Lists two open-source libraries for detecting prompt injection and validating LLM output structure.
Security Guardrails & Policy
Defines mandatory rules for offensive and defensive skills, including disclaimers, user confirmation, and safe design requirements.
llm-guardrails
Introduces Granite Guardian models for detecting harmful prompts and responses, blocking risky queries before they reach an LLM.
AI Safety & Guardrails for Voice Assistants
Defines a multi-layer safety architecture for voice assistants, covering input filtering, deterministic FAQ routing, RAG-grounded AI responses, and output guardrails.
Safety & Guardrails
Defines five kill-switch levels, a cascade prevention system, governance gates, and rollback mechanisms for an autonomous AI coding agent.