Research

Welcome to the 0din Research Hub, a dedicated space for exploring the cutting edge of AI safety and security. Here, we delve into the methodologies, taxonomies, and scoring systems that underpin our understanding of AI risks and mitigations. This section provides access to key resources designed to inform and empower researchers, developers, and policymakers in the responsible development of artificial intelligence.

Security Boundaries

Gain insights into the evolving landscape of AI security with our 2025 Security Boundaries. This document outlines key attack vectors and vulnerabilities, providing a detailed overview of potential threats such as prompt injection, interpreter jailbreaks, and content manipulation. Understand the severity and impact of each threat, along with relevant references and mitigation strategies.

Jailbreak Taxonomy

Explore our comprehensive jailbreak taxonomy, a structured framework for categorizing and understanding various techniques used to bypass AI guardrails. This resource provides a hierarchical view of jailbreak methods, from broad categories to specific techniques, aiding in the identification and mitigation of potential vulnerabilities.

Social Impact Score (SIS)

The Social Impact Score (SIS) is a crucial tool for assessing the potential societal harm resulting from AI violations. This scoring system categorizes risks into five levels, ranging from minimal to critical, based on the severity of potential consequences. Use the SIS to evaluate the ethical implications of AI outputs and prioritize safety measures.

Nude Imagery Rating System (NIRS)

The Nude Imagery Rating System (NIRS) offers a structured approach to classifying nude imagery based on artistic intent, realism, and potential for offense. This system categorizes images into five levels, from abstract representations to explicit content, providing a framework for content moderation and ethical considerations in AI-generated visuals.