AI ETHICS & COMPLIANCE

1. INTRODUCTION AND SCOPE

1.1 Purpose

Legalica OÜ ("Legalica") is committed to developing and deploying artificial intelligence systems that are ethical, transparent, fair, and accountable. This AI Ethics & Compliance Policy establishes the principles and practices governing our use of AI technologies.

1.2 Legal Framework

This Policy is designed to comply with:

1.3 Scope

This Policy applies to:

• (a) All AI systems deployed within the Platform;
• (b) All Third-Party LLMs integrated into the Platform;
• (c) All employees, contractors, and subprocessors involved in AI development;
• (d) All training data, fine-tuning processes, and model outputs;
• (e) All automated systems across all 324 Supported Jurisdictions and 187+ languages.

1.4 Governance Structure

AI Ethics function: AI ethics at Legalica is currently exercised as an internal management function responsible for:

• Reviewing AI system design and deployment decisions;
• Conducting ethics impact assessments;
• Investigating bias reports and complaints;
• Ensuring compliance with this Policy and applicable law.

A formal AI Ethics Committee with external advisors is planned as the company grows. Contact for all AI ethics matters: legal@legalica.app

2. SEVEN CORE ETHICAL PRINCIPLES

3. AI SYSTEM CLASSIFICATION

3.1 Platform AI Components

3.2 Classification Under EU AI Act

3.3 What Legalica Does NOT Deploy

Legalica does NOT deploy any AI systems classified as:

• High-risk AI systems under Annex III (autonomous legal decisions);
• AI systems for social scoring;
• Subliminal techniques to distort behavior;
• Exploitation of vulnerabilities of specific groups;
• Real-time biometric identification in public spaces;
• Emotion recognition in workplace/educational settings;
• AI systems assessing criminal offense risk of natural persons;
• AI systems manipulating human behavior contrary to EU values.

4. DATA GOVERNANCE FOR AI SYSTEMS

5. BIAS DETECTION AND MITIGATION

5.1 Risk Areas

5.2 Testing Protocol

Items marked "target" or "planned" reflect our committed roadmap as an early-stage company and will be updated as processes become fully operational.

5.3 Reporting

Report suspected bias to: legal@legalica.app (subject: "AI Bias Report"). Investigation within 14 business days. Findings documented and disclosed in annual report where appropriate.

6. HUMAN OVERSIGHT FRAMEWORK

6.1 Human-in-the-Loop

User Query → AI Processing → Human Review Gate → Verified Output

The Platform is designed so that AI-generated legal content passes through a human review step before any consequential use. By design, AI-generated legal documents are not:

• Automatically filed with courts or tribunals;
• Sent to third parties without a human action by the user;
• Presented as final legal advice — professional verification is required (see our Terms of Service and Legal Disclaimer).

6.2 Competency Requirements

Users must: (a) possess sufficient legal knowledge to evaluate AI outputs; (b) understand AI capabilities and limitations; (c) exercise independent professional judgment; (d) comply with professional conduct rules.

6.3 Override Mechanisms

Users may: disregard AI suggestions; request alternative outputs; flag incorrect content; escalate to human support.

7. ENVIRONMENTAL IMPACT

7.1 Environmental Goals

The following are aspirational targets that we monitor as the company grows. Actual figures depend significantly on our infrastructure and AI providers.

7.2 Efficiency Measures

Intelligent model routing (smaller models for simpler tasks); intelligent caching; batched processing; energy-efficient architectures.

8. REPORTING AND AUDITS

8.1 Annual Ethics Report

First report planned: Q1 2027. Covers: bias audits, environmental metrics, complaint statistics, policy updates, third-party findings.

8.2 Audit Schedule

8.3 Complaint Mechanism

Email: legal@legalica.app (subject: "AI Ethics Complaint")

Postal: Legalica OÜ, Attn: AI Ethics, Ahtri tn 12, 10151 Tallinn, Estonia. Acknowledged within 5 business days. Investigated within 30 business days. Kept confidential.

9. POLICY GOVERNANCE

9.1 Review: Annually or upon: material AI system changes, legal changes, material ethics incidents, management request.

9.2 Training: All employees/contractors in AI development: ethics training upon onboarding + annual refresher + written acknowledgment.

10. REGULATORY CLASSIFICATION

10.2 GPAI Model Obligations: Legalica acts as a deployer of general-purpose AI (GPAI) models provided by third parties (Google Vertex AI / Gemini, Moonshot AI / Kimi, Groq / Llama, and local open-source models). The GPAI provider obligations under Articles 53–55 of the AI Act (technical documentation, copyright policy, training data summaries, and — for models with systemic risk — model evaluation, systemic-risk assessment, and serious incident reporting) rest with those model providers. Legalica does not substantially modify GPAI models and does not rebrand them under its own name; should this change, Legalica will assume the corresponding provider obligations under Article 25.

11. ARTICLE 50 TRANSPARENCY OBLIGATIONS

The transparency obligations of Article 50 of the AI Act are applicable since 2 August 2026 and are implemented as follows:

11.1 User Information

11.2 How Users Are Informed

(a) the Platform interface itself, which expressly presents every assistant interaction as AI-powered; (b) this AI Ethics & Compliance Policy at legalica.app/trust/ai-ethics; (c) Terms of Service Section 5 (AI Act Compliance); (d) Legal Disclaimer prominently displayed; (e) Privacy Policy Section 12.

12. RISK MANAGEMENT SYSTEM

12.1 Risk Assessment: Risk assessments evaluate: (a) health, safety, and fundamental rights risks; (b) intended use and foreseeable misuse risks; (c) specific group risks; (d) bias and discrimination risks; (e) data quality risks.

12.2 Mitigation:

12.3 Incident Reporting: Report to: legal@legalica.app (subject: "AI Incident"). Investigation within 14 business days.

13. TECHNICAL DOCUMENTATION

Legalica maintains documentation including: (a) general AI system description; (b) design specifications; (c) data requirements; (d) known limitations; (e) performance metrics; (f) deployment procedures.

Available to regulatory authorities upon request and to Enterprise customers under NDA.

14. RECORDS AND TRANSPARENCY REPORTING

14.1 Records: AI system versions and deployments; reference data sources; risk assessments; incident reports; user feedback; audit results.

14.2 Transparency Reports: Annual reports covering: AI capabilities/limitations, bias audit results, incident statistics, user feedback analysis, improvements. First report planned: Q1 2027.

15. PROHIBITED AI PRACTICES (AI ACT ARTICLE 5)

Legalica confirms it does not and will not:

• Deploy subliminal techniques to distort behavior causing harm;
• Exploit vulnerabilities of specific groups causing harm;
• Use AI for social scoring by governments;
• Use real-time remote biometric identification in publicly accessible spaces for law enforcement;
• Assess risk of natural persons committing criminal offenses;
• Create or expand facial recognition databases through untargeted scraping;
• Infer emotions in workplace or educational institutions.

16. MULTI-JURISDICTIONAL AI COMPLIANCE

16.1 Global AI Regulation Landscape: Legalica monitors AI regulation development across all 324 Supported Jurisdictions. Key regulatory frameworks:

16.2 Compliance Strategy: (a) EU AI Act serves as baseline standard; (b) GDPR alignment for all AI processing; (c) proactive monitoring of regulatory changes across 324 jurisdictions; (d) jurisdiction-specific adaptation; (e) service availability limits under Force Majeure if operation becomes non-viable in a jurisdiction.

16.3 Local AI Ethics Requirements: Some jurisdictions have specific AI ethics requirements: (a) EU: AI Act mandatory requirements; (b) Germany: BDSG AI provisions; (c) France: CNIL AI guidance; (d) Singapore: PDPA AI framework; (e) South Korea: AI Basic Act mandatory requirements.

17. LANGUAGE-SPECIFIC AI GOVERNANCE

17.1 Fairness Across Languages: The Platform supports 187+ languages. AI ethics principles apply equally across all languages: (a) no language discrimination; (b) transparency notice in user's language; (c) bias metrics tracked per language; (d) human review requirements apply universally.

17.2 Low-Resource Language Ethics: For low-resource languages: (a) users are informed of potential accuracy limitations; (b) additional human verification is recommended; (c) continuous investment in low-resource language support; (d) feedback collection.