AI Transparency

Last updated:

1. Introduction

In accordance with the EU AI Act (Regulation 2024/1689), Article 50, Scudo AI provides this transparency disclosure about our artificial intelligence systems and practices.

Risk Classification: Our AI systems are subject to the transparency obligations of Article 50 of the EU AI Act and, in standard deployments, are not classified as high-risk. High-Risk classification under Article 6 and Annex III may apply per deployment for relevant use cases (including healthcare, employment, credit-scoring, critical infrastructure, law enforcement, biometrics) and is determined in each Order Form.

2. Our AI Systems

Scudo AI develops and deploys the following AI systems:

Sovereign Business Intelligence Platform

Purpose
Builds private knowledge graphs from customer documents, supports entity extraction, cross-document search and analytics
Deployment
On-premises
Human Oversight
Yes

AI Compliance Testing System

Purpose
Evaluates large confidential document sets against customer compliance rules, produces a full decision trail for every flagged item and routes it for human review
Deployment
On-premises
Human Oversight
Yes

Desktop Process Automation Platform

Purpose
Runs local business process automation on operator workstations with local LLM document extraction, classification and routing
Deployment
On-premises
Human Oversight
Yes

Voice Transcription System for Regulated Work

Purpose
Transcribes clinical, forensic and legal speech locally in Polish and English, with source-audio linkage for traceability
Deployment
On-premises
Human Oversight
Yes

3. Core Principles

Local Processing by Default

Designed for fully local processing - in standard on-premises deployments, customer business data is processed on customer-controlled hardware. Specific data flows are defined in each Order Form and DPA.

Customer Data Ownership

You retain ownership of customer data you provide; output rights are governed by the Order Form.

Human Oversight

All AI recommendations are subject to human review and approval.

4. AI Capabilities and Limitations

Capabilities

  • Natural language processing for domain-specific tasks
  • Document analysis and information extraction
  • Pattern recognition and anomaly detection
  • Recommendation generation based on input data

Limitations

  • Our AI provides recommendations, not definitive decisions
  • Performance depends on quality and completeness of input data
  • AI may produce incorrect or incomplete outputs
  • Models are trained for specific use cases and may not generalize
  • Human oversight is always recommended for critical decisions

5. Human Oversight

All Scudo AI systems are designed with human oversight as a core principle:

  • Human-in-the-loop: All AI recommendations require human review and approval
  • Override capability: Humans can always override AI decisions
  • Audit trails: All AI actions are logged for review
  • Training: We provide training for users on AI capabilities and limitations

6. Data and Privacy

Our Private AI architecture is designed to maximize privacy:

  • Local Processing by Default: In standard on-premises deployments, data is processed on customer-controlled hardware
  • No Cloud Transmission in standard configuration: Data stays on your premises by default
  • Air-gapped Operation: Systems are designed to operate without internet
  • Customer Data Ownership: You retain ownership of customer data you provide; output rights are governed by the Order Form

We do not fine-tune on customer data without explicit written consent. Base models we deploy are sourced from publicly available datasets and/or datasets used under applicable law and licenses available to the model provider; we maintain a model card per system.

7. Bias and Fairness

We aim to identify and reduce bias in AI systems:

  • Bias-testing scope is defined per deployment
  • Diverse training data where applicable
  • Human review of AI recommendations
  • Continuous monitoring and improvement

If you observe potentially biased behavior in our AI systems, please report it to: office@scudoai.com

8. Contact for AI-Related Inquiries

For questions about our AI systems, transparency practices, or to report concerns:

Email: office@scudoai.com
Subject: AI Transparency Inquiry