AI Transparency & Disclosure

NOTICE: EU AI Act Compliance (Regulation (EU) 2024/1689)

This page provides required transparency disclosures for artificial intelligence systems offered by MYG Media SRL in compliance with the European Union Artificial Intelligence Act. All AI systems deployed are clearly labeled, their capabilities and limitations disclosed, and human oversight mechanisms in place.

1. AI Systems We Use

MYG Media SRL integrates artificial intelligence into multiple services and products. This page provides complete transparency about:

• Which AI systems we use (third-party and custom)
• How these systems work and their limitations
• What data they process
• Human oversight and control mechanisms
• Your rights when interacting with AI

2. Third-Party AI Services

We integrate the following third-party AI systems to deliver our services. Important: MYG Media SRL does not control these systems and is not responsible for their outputs, errors, biases, or failures.

3. Custom Software vs. Third-Party Solutions

3.1 Custom Software Development

• Proprietary code written specifically for your business
• Full control over functionality and data processing
• Custom AI training on your specific data
• Higher cost, longer development time
• Source code ownership negotiable per contract

3.2 Third-Party Service Integration

• Integration of existing SaaS platforms and APIs
• Faster deployment, lower cost
• Subject to third-party terms, pricing, and availability
• MYG Media SRL provides configuration, not underlying software
• Third-party provider controls updates, features, uptime

3.3 Hybrid Approach (Most Common)

Custom logic and workflows built on top of third-party AI APIs. Example: Custom-built chatbot interface using OpenAI GPT-4 API for intelligence. You own the custom code, third-party provides AI capabilities.

4. High-Risk AI Systems

Our Trade Intelligence AI system assists with trade document validation and compliance checking. Under EU AI Act Article 6, systems that may impact:

• Access to essential public or private services (trade/customs)
• Legal/regulatory compliance decisions

may be classified as high-risk. In accordance with the AI Act:

• Human Oversight: All Trade Intelligence outputs require human review before action
• Not a Replacement for Professionals: AI assists; humans decide
• Accuracy Monitoring: System accuracy tracked and reported quarterly
• Bias Mitigation: Regular testing for discriminatory outputs
• Right to Human Review: Users can request human expert review of any AI decision
• Transparency: AI confidence scores shown for all outputs

5. AI System Limitations

All AI systems, both third-party and custom-built, have inherent limitations:

5.1 Hallucinations (False Information)

AI language models can generate plausible-sounding but factually incorrect information. This occurs unpredictably and cannot be fully prevented. Users must verify all AI-generated content before use.

5.2 Training Data Cutoff

Large language models are trained on historical data with a cutoff date. They do not have access to real-time information, current events, or data beyond their training cutoff unless explicitly provided by users or through integrations.

5.3 Bias & Discrimination

AI models may exhibit biases present in their training data (gender, racial, cultural, linguistic). We implement bias detection and mitigation, but cannot guarantee bias-free outputs. Report discriminatory outputs immediately to max@myg-media.com.

5.4 No Guaranteed Accuracy

MYG Media SRL provides NO WARRANTY regarding AI output accuracy. While we strive for high quality:

• AI may produce incorrect, incomplete, or misleading information
• Accuracy varies depending on task complexity and input quality
• Edge cases and novel scenarios may produce unreliable results
• Third-party AI providers control model quality, not MYG Media SRL

6. Human Oversight & Control

All MYG Media SRL AI systems include human oversight mechanisms:

• Human-in-the-Loop: Critical decisions require human approval
• Override Capability: Humans can override or reject AI suggestions
• Escalation Path: Uncertain AI outputs flagged for human review
• Audit Trails: All AI interactions logged for review
• Manual Fallback: Systems function without AI if needed
• Emergency Stop: AI processing can be halted immediately

7. Your Rights When Interacting with AI

8. AI Training & Your Data

8.1 Third-Party AI Models

OpenAI, Anthropic, and Google have their own data policies:

• OpenAI API: Data sent via API is NOT used for model training (as of March 2023 policy)
• Anthropic Claude: Data retention policies vary by contract tier
• Google Gemini: Enterprise customers can opt out of data retention

We configure these services with maximum privacy settings available. However, MYG Media SRL cannot control third-party data handling beyond their published policies.

8.2 Custom AI Training

When building custom AI systems:

• Training data is isolated per client (not shared across projects)
• Client data used for training only with explicit written consent
• Training data retention periods specified in service agreement
• Clients can request deletion of all training data at any time
• No personally identifiable information (PII) used for training without anonymization

9. Reporting AI Issues

If you encounter any of the following, report immediately:

• Discriminatory, biased, or offensive AI outputs
• Factually incorrect information that could cause harm
• Privacy violations (AI exposing other users' data)
• Security vulnerabilities in AI systems
• Unexpected or dangerous AI behavior

10. Updates to This Disclosure

This AI Transparency & Disclosure page is updated as we add new AI systems or change existing implementations. Material changes are announced via email to active clients and posted on our website for 30 days before taking effect.

11. Contact & Further Information

This AI Transparency & Disclosure page complies with EU AI Act (Regulation (EU) 2024/1689), GDPR (Regulation EU 2016/679), and applicable Romanian law.