Ethical AI Use Declaration

Ethical AI Use Principles at Viking Potatoes sp. z o.o.

1. Purpose and Legal Basis

Artificial Intelligence (AI) is a breakthrough technology with many positive applications. Aware of its impact on employees, clients and society, Viking Potatoes sp. z o.o. introduces this Declaration to ensure responsible, transparent and ethical use of AI in our business.

The Declaration is based on the following legal acts and initiatives:

Regulation (EU) 2024/1689 of the European Parliament and of the Council of 13 June 2024 laying down harmonised rules on artificial intelligence (the AI Act) - entered into force on 1 August 2024, with phased implementation through 2027.

Article 50 of the AI Act - transparency obligations for users of generative AI systems (Viking Potatoes informs whenever content is generated by AI).

Regulation (EU) 2016/679 (GDPR) - regarding the processing of personal data using AI.

OECD Recommendations on Trustworthy AI, defining key concepts and ethical values (transparency, accountability, non-discrimination).

2. AI Definitions Consensus

In line with international standards, we adopt the following definitions:

AI System: a machine-based system that can make predictions, recommendations or decisions affecting the environment in which it operates.

AI System Lifecycle: phases of design, verification, deployment and monitoring of an AI system, conducted iteratively.

AI Knowledge: resources (data, algorithms, models and know-how) necessary to manage AI.

AI Actors: individuals and organisations participating in the AI system lifecycle.

Stakeholders: entities involved in or affected by the operation of AI systems.

3. Specific AI Tools Used by Viking Potatoes

In our daily work we use the following AI tools:

ChatGPT (OpenAI, USA) - generation of marketing content, product descriptions, text drafts. Individual account (ChatGPT Plus/Pro) with the "Improve the model for everyone" option disabled in privacy settings.

Claude (Anthropic, USA) - document analysis, text editing, code review, project work. Claude Max plan - by default Anthropic does not train its models on consumer-plan user data.

Perplexity Pro (Perplexity AI, USA) - searching for information with up-to-date internet context, competitive research. Pro account with the "AI Data Retention" option disabled.

Gemini (Google LLC, USA) - integration with Google Workspace, SEO data analysis, multimodal content analysis. "Gemini Apps Activity" disabled in My Activity Google to prevent conversations from being used for model training.

Bielik (Polish open-source model, SpeakLeash) - processing of sensitive data in Polish, initial pre-processing of client materials. Runs locally on Viking Potatoes infrastructure - no data transfer outside our infrastructure.

We do not use Enterprise/Team plans - we rely on privacy settings of individual accounts and the locally hosted Bielik model for sensitive data. We verify these settings at least quarterly.

4. Protection of Client Data in AI Processes

When working with client data using AI tools, Viking Potatoes follows these principles:

Training opt-out for our data: in every cloud tool we manually enable privacy settings that disable the use of conversations for model training. Under the Claude Max plan no training on consumer-plan data takes place by default.

Bielik locally for sensitive data: for particularly sensitive client data we use the Bielik model running locally on Viking Potatoes infrastructure - without transfer to external AI providers.

Pseudonymisation: before being entered into AI tools, personal data (names, emails, tax IDs, addresses) is pseudonymised or anonymised whenever its full form is not necessary for the task.

No special category data in cloud AI: special category data within the meaning of Article 9 GDPR (medical, religious, sexual, biometric) is not transferred to cloud AI tools. Where work on such data is necessary, only the local Bielik model is used, or the work is performed without AI.

Review procedure: before any AI output reaches the client, it is verified by a team member for accuracy, fit, hallucinations and brand alignment.

Client right to information: the client may at any time request information about which parts of the delivered material were generated by AI and which were created by a human.

5. Scope of AI Use

We use AI in the following areas of our business:

Content generation and editing (text, graphics, code) - within client projects and internal materials.

Local marketing - data analysis, optimisation of Google Ads and Meta Ads campaigns.

Support of the website design process - prototypes, layout suggestions, UX optimisation.

Automation of internal processes - handling of repetitive administrative tasks.

6. Informing Clients and Obtaining Consent

Our collaboration agreements include a clause regarding the use of AI, informing the client about how it is applied in project delivery. The use of AI beyond standard process support requires the client’s explicit consent. The client may at any time decline the use of AI in their project.

7. Quality Control of AI Outputs

All content and results generated by AI undergo multi-stage verification before being delivered to the client:

Fact verification - checking that the AI has not generated fictional data, statistics, quotes or references (hallucination control).

Brand voice verification - alignment with the client’s communication style and brand guidelines.

Copyright verification - checking that the generated content does not substantially copy protected works of third parties.

Team member acceptance - every piece of content, before being sent to the client, requires acceptance from an experienced team member (not the AI operator themselves).

Right to complaint - the client has the right to raise objections regarding the quality of AI-generated content within 30 days from the date of its delivery. Complaints are handled in accordance with the procedure set out in the Terms of Service.

8. Security and Risk Management

Continuous monitoring: we monitor AI systems to identify and eliminate potential errors.

Risk matrices: detailed risk assessment for each AI application.

Escalation procedures: clearly defined paths for reporting and resolving issues.

Security protocols: detailed data protection procedures.

Retention rules: defined data retention periods.

Backup procedures: regular backups and data recovery procedures.

9. Ethics and Social Responsibility

Our AI activities are based on:

Supporting, not replacing human work: AI is a supporting tool, and the human remains the decision-maker.

Continuous improvement and education: we update our procedures to keep pace with technological and legal changes.

10. Known Limitations of AI Tools

Despite multi-stage quality verification, AI tools have known limitations of which we inform clients:

Hallucinations - AI models may generate information that sounds credible but is untrue. We always verify facts before delivery.

Bias - AI models may reflect biases stemming from their training data. We check content for unintended stereotypes or oversimplifications.

Knowledge cutoff - AI models have a defined training "cutoff" and may not be aware of the latest events or changes. We verify timeliness against external sources.

Copyright - intellectual property of AI-generated content is evolving in law. Currently in PL/EU the rule applies that AI on its own does not create works within the meaning of copyright law (no human authorship), but our editors may creatively modify the texts.

11. Controller Data and Contact

Viking Potatoes sp. z o.o., ul. Krypska 25/9, 04-082 Warsaw, Poland. KRS 0000648228, NIP 5862312148, REGON 365907852. Contact for AI-related matters: ratunku@chcedointernetu.pl .

Last updated: 12 June 2026 (Version 1.1).

Appendices

Appendix 1 - General AI Use Procedure - Read the AI use procedure

Appendix 2 - Monitoring - Read the monitoring procedure