training-data-review

What problem does it solve?

This Skill solves the critical problem of unregulated AI/ML training data usage that violates Russian federal laws, exposing organizations to legal penalties, model liability, and operational risks for non-compliance with personal data, copyright, and sanctions rules.

Core Features & Use Cases

• Comprehensive Compliance Framework: Provides structured checklists for 152-FZ personal data legal basis assessment, anonymization guidance, and special category data handling tailored to Russian regulations.
• Copyright & IP Risk Mitigation: Covers Russian copyright rules (no US-style fair use doctrine), licensing requirements for public, purchased, and scraped data, and high-risk source identification.
• End-to-End Audit Workflow: Guides users through a 5-step process from training data inventory to remediation planning and formal datasheet documentation, with built-in risk scoring for data sources.
• Use Case: A team fine-tuning a Russian-language customer service LLM can use this Skill to audit their internal chat logs, public Wikipedia dumps, and licensed news datasets to identify 152-FZ and copyright gaps before training begins.

Quick Start

Invoke the training-data-review skill to conduct a full compliance audit of your planned AI model training dataset against Russian personal data, copyright, and sanctions regulations.