Artificial Intelligence Policy
Ethical, secure, and transparent use of AI technologies in our training bootcamps, workshops, and enterprise consulting.
Applicable Regulations:
- Regulation (EU) 2024/1689 of 13 March 2024 (EU AI Act)
- Regulation (EU) 2026/1744 of 24 July 2026 (Amendment)
- Regulation (EU) 2016/679 (GDPR) - Arts. 13, 14, 22
- Regulation (EU) 2023/2854 (Cyber Resilience Act)
Scope: All Matrits educational offerings (bootcamps, corporate courses, coaching sessions, webinars)
Last updated: September 2026 · Version 2.0 (revised for compliance with amendments 2026/1744)
Plain Language Summary
- Active Pedagogy: We teach real-world use of foundational models (Anthropic Claude, OpenAI ChatGPT) for workflow automation and productivity.
- Data Confidentiality: We never train public AI models with your personal or proprietary business data. We use enterprise sandboxes with zero retention.
- Human-in-the-Loop: Every course is designed, verified, and delivered by human senior instructors. No automated evaluations determine course outcomes.
- EU Regulatory Alignment: All systems and methodologies adhere strictly to European safety, transparency, and ethics regulations.
Four Fundamental Principles
1. Transparency & Disclosure
Participants are always informed when an AI tool is being demonstrated, how prompts are processed, and where data travels.
2. Data Privacy & Zero-Training
We configure commercial accounts and Claude Projects so that client interactions are never used by providers to train future models.
3. Ethical & Critical AI
We teach rigorous hallucination detection, prompt constraints, validation methods, and bias mitigation in corporate workflows.
4. Strict Compliance (EU AI Act)
Our educational frameworks comply with Articles 50 & 52 of the European AI Act regarding general purpose AI and transparency.
1. Scope and Application
This policy applies to all training programs, bootcamps, workshops, executive reskilling sessions, and educational materials organized by Matrits IA, S.L. (hereinafter 'Matrits' or 'the Company'), whether conducted online or onsite.
2. AI Systems & Providers in the Classroom
During our sessions, students learn to operate leading commercial LLMs (Large Language Models) under enterprise best-practice guidelines:
Used for deterministic document analysis, operational automation, and structured prompt engineering. Commercial privacy policies apply with zero training on user prompt data.
Used for comparative analysis, data manipulation, and productivity workflows with privacy controls activated.
Where high-confidentiality workflows are taught, we demonstrate local offline processing where zero bytes leave the user's computer.
Editorial Transparency in AI-Generated Content
When Matrits publishes content, exercises, pedagogical materials, or public interest texts developed with the assistance of AI systems (Claude, GPT-4o), we explicitly identify:
- Which components were generated by AI (structure, drafts, analysis)
- What human review was applied (proofreading, validation, editorial editing)
- Who assumes editorial responsibility (tutor, senior consultant, content lead)
3. Student Data Protection & Practical Cases
We enforce strict privacy hygiene during all live workshops and assignments:
- Synthetic & Anonymized Data: Exercises use fictional company datasets, synthetic financial sheets, and dummy records to prevent real sensitive exposure.
- In-Company Custom Audits: When handling real enterprise workflows, Non-Disclosure Agreements (NDA) and private corporate sandboxes are established prior to class.
- Recordings & Material Sharing: Class recordings are accessible only by registered participants through authenticated student portals and are never shared publicly.
3b. Automated Decisions, Profiling & AI Systems (Arts. 13, 14, 22 GDPR)
At Matrits we recognize that certain processes may involve automated processing of personal data or profiling pursuant to Articles 13, 14, and 22 of the General Data Protection Regulation (GDPR).
When Do We Use AI on Your Data?
We utilize artificial intelligence systems in the following specific contexts:
- Analysis of evaluations and practical assignments: We use Claude or GPT-4o to provide initial automated feedback on projects, code, or analysis submitted by students. This feedback is ALWAYS reviewed by a human tutor before being communicated.
- Generation of pedagogical recommendations: When requested by the student, we deploy AI systems to suggest personalized learning paths based on their exercise responses. This recommendation is informative and non-binding.
- Data analysis in corporate courses: In in-company training where the client authorizes analysis of aggregated organizational data, we employ AI to extract patterns or trends. These datasets are ALWAYS anonymized and public models are never trained on them.
Information Regarding the Logic Applied (Art. 22.3 GDPR)
For each automated decision or profiling process, Matrits provides:
- The model (Claude/GPT-4o) is trained on data up to a specific cutoff date.
- It processes your input (exercise, question, data) and generates responses based on statistical patterns.
- It does NOT use prior decisions about you to bias future results (no dynamic fine-tuning or retraining).
- It can produce inaccurate output or hallucinations (false claims delivered with high confidence).
- It may reflect biases present in training data (particularly in demographic, gender, and origin domains).
- It has no access to information subsequent to its training cutoff.
- It lacks independent legal, ethical, or normative reasoning capabilities.
- If feedback is inaccurate, you can request immediate human review at zero cost.
- If a pedagogical suggestion does not fit your goals, your tutor can adapt or cancel it without administrative friction.
- If corporate data is used, the client receives aggregated pattern insights, never individualized profiling on staff members.
Your Right Not to Be Subject to Automated Decisions (Art. 22 GDPR)
You have the right not to be subject to a decision based solely on automated processing producing legal or similarly significant effects. At Matrits:
- No student is suspended, expelled, or certified based solely on automated grading. Final grading decisions always originate from a human instructor.
- We do not deploy AI to grade exam submissions or capstone projects without prior human assessment. The system may suggest preliminary scores, but only a certified consultant can grant credits or diplomas.
- You retain the right to request human intervention at any stage. If you disagree with AI-generated feedback, contact your tutor for a direct manual review.
How to Request Review, Explanation, or File a Complaint
If an automated decision affects you significantly (particularly in evaluation, certification, or employment references):
- Request an explanation: Contact info@matrits.com referencing the output. We will provide full transparency on how data was processed and why that output was returned.
- Request human review: A senior consultant will perform manual evaluation of your case, considering contextual factors an automated model might overlook.
- State your viewpoint: Submit additional context or arguments you wish considered (e.g., specific pedagogical framing of code that was undetected).
- Contest the decision: If post-review you still consider the decision flawed, you may escalate to: 1) Matrits AI Governance Officer (10 business days response window), and 2) Your national Data Protection Authority (AEPD in Spain).
Confidentiality: We Do Not Train Models With Your Data
Matrits explicitly commits to:
- Never sell, share, or transfer personal data to AI vendors: Neither OpenAI, Anthropic, nor any third party has access to student names, emails, exercises, or corporate data without specific consent.
- Disable training history: When using Claude Projects or ChatGPT at Matrits, interactions are never used to train public foundation models. Commercial enterprise zero-retention policies are configured.
- Anonymize data for aggregate metrics: When reporting aggregate training success metrics (e.g., '75% of participants streamline office workflows post-bootcamp'), personal identifiable data is strictly excluded.
4. Prohibited Practices & Non-Negotiables
Matrits explicitly commits NEVER to:
- Sell or license student data to AI vendors or brokers.
- Automate grade assignments or student certifications without human oversight.
- Deploy non-consensual biometric or emotion recognition systems.
- Encourage the bypass of copyrighted or unauthorized scraping models.
5. Intellectual Property of Student Creations
All prompt libraries, automated workflows, custom tools, and systems created by students during our bootcamps remain the sole intellectual property of the student or their employer. Matrits claims no ownership over projects built by participants.
6. Human Oversight Protocol & EU AI Act Compliance Matrix
6.1 Detailed Human Oversight Protocol (Arts. 14 & 26 EU AI Act)
Pursuant to Articles 14 and 26 of Regulation (EU) 2024/1689, Matrits implements continuous and effective human oversight over all AI systems deployed in training sessions, bootcamps, and corporate programs.
Responsibility Structure by AI System
| AI System / Function | Responsible Supervisor | Intervention Capacity | Response Time | Stop Protocol |
|---|---|---|---|---|
| Document analysis (Claude Projects) | Assigned Senior Tutor/Consultant | Immediate session halt + output review | Real-time (< 2 min) | Notify student + request manual content review |
| Content generation (Claude/GPT-4o) | Lead Pedagogical Designer | Pre-delivery review + classroom pre-check | Pre-session (24h before) | Material rewrite or deletion |
| Workflow automation (n8n/Zapier) | Matrits Technical Lead | Workflow deactivation + manual intervention | 30 minutes max | Incident log + workflow security audit |
| Automated exercise correction | Senior Consultant + Student | 100% human review prior to final grading | Prior to grade communication | Mandatory personalized feedback |
How Supervision Works in Three Phases
1. Before the session: All materials generated with AI assistance undergo review by a senior consultant ≥ 48 hours prior to delivery.
2. During the session: The tutor monitors all interactions with AI systems, is present in person (onsite courses) or virtually (live remote sessions with recording), and possesses immediate authority to halt any exercise if inaccurate, biased, or potentially harmful outputs occur.
3. After the session: Interaction logs are reviewed, any incident or anomaly is documented, and findings are communicated to the AI Governance Officer.
Effective Capacity to Intervene
Human oversight is never merely passive. Matrits tutors and consultants possess:
- Decision authority: They can halt any AI usage immediately without prior administrative authorization.
- Technical competence: Every supervisor is trained in limitations, hallucinations, biases, and model failure modes.
- Access to system controls: They hold credentials to deactivate, pause, or reconfigure any AI system in active use.
- Documented accountability: A comprehensive registry of all interventions, decisions, and justifications is maintained.
Review of Outputs Prior to Producing Significant Effects
Pursuant to Art. 26.2.d of the EU AI Act, AI outputs are systematically reviewed before producing significant outcomes:
- In practical exercises: The tutor reviews the generated response/output before authorizing the student to incorporate it into their deliverable.
- In diagnostics: Any analysis or recommendation generated by AI is benchmarked against human expertise before presentation to corporate clients.
- In diplomas/certifications: They are NEVER issued automatically. A senior consultant manually verifies that the student satisfies all evaluation criteria.
- In progress reports: AI-assisted progress reports invariably incorporate qualitative human observations from a lead instructor.
6.2 EU AI Act Compliance Matrix
| Requirement | Classification | Matrits Implementation |
|---|---|---|
| Art. 50 (Transparencia) | Obligatorio | Clear labeling of all AI outputs and live demos. |
| Art. 52 (Información al usuario) | Riesgo Limitado | Detailed explanation of model parameters and limitations. |
| Art. 14 (Supervisión humana) | Garantizado | 100% human-guided syllabus and tutoring. |
| RGPD / GDPR (Privacidad) | Blindado | Opt-out from training pipelines + zero data retention. |
7. Student Rights & Contact Channel
Any participant may request detailed information regarding the AI systems utilized in their program, as well as submit inquiries or complaints concerning this policy.
8. Matrits Team AI Literacy Program
Matrits develops and implements formal AI literacy programs for its staff:
- Foundational model operations (Claude, GPT-4o): logic, biases, limitations
- Human oversight and hallucination detection (hands-on with real cases)
- Compliance: GDPR, EU AI Act, user rights
- AI ethics and practical use cases in education and consulting
Quarterly update minimum (models evolve, regulations change).
José Antonio Díaz Infante, AI Governance Officer.
Attendee logs, dates, and syllabus (internal; available upon GDPR request).
Evidence: See diploma.jpg — Matrits team training certificate.
Explicit Commitment to Continuous Training
Matrits commits to:
- Equip each team member with initial AI training (minimum 8h, first month)
- Quarterly refresh sessions on emerging models, regulatory revisions, and use cases
- Access to self-learning resources (Anthropic docs, OpenAI guides, research papers)
- Annual competency assessment (not to certify a specific tier, but to identify gaps)
- Designated Officer: José Antonio Díaz Infante — answers any inquiry regarding competency or methodology
No mandatory minimum level (Art. 4 amended July 2026), but a documented commitment to "reasonable measures".
9. AI Governance Officer
José Antonio Díaz Infante, AI Governance and Regulatory Compliance Officer, is the identifiable person responsible for the oversight, implementation, and review of this Artificial Intelligence Policy and all compliance measures pursuant to Regulation (EU) 2024/1689, as amended by (EU) 2026/1744.
Direct Contact for AI Governance Inquiries
- Email: info@matrits.com
- Availability: Guaranteed response within a maximum of 2 business days
Any matter concerning governance, risk classification, human oversight, or amendment of this policy should be addressed to this direct contact.
AI Ethics and Transparency Desk — Available on Two Levels
Both channels are confidential and documented.
For inquiries regarding governance, human oversight, and regulatory compliance:
- Email: gobernanza-ia@matrits.com
- Officer: José Antonio Díaz Infante
- Response time: 2 business days
For general inquiries regarding training, cookies, or privacy:
- Email: joseantonio@matrits.com
- Response time: 3 business days
Both channels are confidential and documented.