CSOAI - Council of Safety of AI

The World's First AI Safety Infrastructure Platform

CSOAI is a comprehensive AI safety governance platform providing multi-framework compliance, Byzantine fault-tolerant decision-making, and workforce development for the AI safety industry.

Overview

CSOAI serves three core functions:

1. Enterprise Compliance - Multi-framework AI compliance automation (EU AI Act, NIST AI RMF, ISO 42001, TC260)
2. 33-Agent Council - Byzantine fault-tolerant voting system for AI safety decisions
3. Training & Certification - Professional development for AI safety analysts

Features

Multi-Framework Compliance

• EU AI Act - 113 articles, full compliance mapping
• NIST AI RMF - 72 requirements across GOVERN, MAP, MEASURE, MANAGE
• ISO 42001 - AI Management System standard
• TC260 - Chinese AI Safety Governance Framework

33-Agent Council (Byzantine Fault Tolerant)

• 33 independent AI agents across 3 providers
• 11 Guardian agents (safety, security, privacy)
• 11 Arbiter agents (fairness, transparency, accountability)
• 11 Scribe agents (documentation, compliance, reporting)
• 22/33 consensus threshold
• Tolerates up to 10 faulty/malicious agents

Training & Certification

• 8-module courses for each framework (24 total modules)
• Professional certification exams
• Analyst workbench for case management
• Job marketplace for AI safety professionals

Watchdog System

• Public incident reporting
• AI model safety leaderboard
• Community-driven accountability

Tech Stack

Project Structure

``` CSOAI/ ├── client/ # React frontend │ ├── src/ │ │ ├── components/ # Reusable UI components │ │ ├── pages/ # Page components │ │ ├── contexts/ # React contexts │ │ ├── hooks/ # Custom hooks │ │ ├── lib/ # Utilities and tRPC client │ │ └── styles/ # CSS styles │ ├── index.html │ └── vite.config.ts ├── server/ # Node.js backend │ ├── routers/ # tRPC routers │ ├── services/ # Business logic │ │ ├── council.ts # 33-Agent Council │ │ ├── auth.ts # Authentication │ │ └── ... │ ├── db/ # Database schema and config │ ├── content/ # Training module content │ └── index.ts # Server entry point ├── shared/ # Shared types and schemas ├── docs/ # Documentation │ ├── architecture/ # Technical architecture │ ├── compliance/ # Compliance frameworks │ ├── courses/ # Training content │ └── business/ # Business docs ├── tests/ # Test files └── assets/ # Static assets ```

Getting Started

Prerequisites

• Node.js 20+
• pnpm
• PostgreSQL or MySQL database
• API keys for OpenAI, Anthropic, and Google AI

Installation

1. Clone the repository ```bash git clone https://github.com/csoai/platform.git cd CSOAI ```

2. Install dependencies ```bash pnpm install ```

3. Configure environment ```bash cp .env.example .env

Edit .env with your configuration

```

4. Set up database ```bash pnpm db:push ```

5. Start development server ```bash pnpm dev ```

The frontend will be available at http://localhost:5173 The API will be available at http://localhost:3001

Available Scripts

• `pnpm dev` - Start development servers
• `pnpm build` - Build for production
• `pnpm start` - Run production server
• `pnpm test` - Run tests
• `pnpm db:push` - Push schema changes to database
• `pnpm db:studio` - Open Drizzle Studio

API Documentation

Authentication

• `POST /api/trpc/auth.login` - User login
• `POST /api/trpc/auth.register` - User registration
• `GET /api/trpc/auth.me` - Get current user

Compliance

• `GET /api/trpc/compliance.getFrameworks` - List frameworks
• `POST /api/trpc/compliance.startAssessment` - Start assessment
• `GET /api/trpc/compliance.getAssessment` - Get assessment

33-Agent Council

• `GET /api/trpc/council.getAgents` - List all agents
• `GET /api/trpc/council.getSessions` - List voting sessions
• `POST /api/trpc/council.triggerVoting` - Trigger council vote

Watchdog

• `GET /api/trpc/watchdog.getReports` - List reports
• `POST /api/trpc/watchdog.submitReport` - Submit report
• `GET /api/trpc/watchdog.getLeaderboard` - Safety leaderboard

Training

• `GET /api/trpc/training.getCourses` - List courses
• `POST /api/trpc/training.enrollInCourse` - Enroll in course

Compliance Frameworks

EU AI Act (2024)

The EU AI Act is the world's first comprehensive legal framework on AI. CSOAI maps all 113 articles and provides:

• Risk classification (Unacceptable, High, Limited, Minimal)
• Conformity assessment workflows
• Technical documentation generation
• Human oversight verification

NIST AI RMF 1.0

The NIST AI Risk Management Framework provides:

• GOVERN: Policies and procedures
• MAP: Risk identification
• MEASURE: Risk assessment
• MANAGE: Risk treatment

ISO/IEC 42001

AI Management System standard covering:

• Leadership and planning
• Risk assessment
• Operational controls
• Performance evaluation

TC260 (China)

Chinese AI Safety Governance Framework:

• Inherent Risk assessment
• Application Risk evaluation
• Derivative Risk monitoring

License

MIT License - see LICENSE file for details.

Contact

• Website: https://councilof.ai
• Email: contact@councilof.ai