How we use AI: European Pravda's internal rules

European Pravda is a newsroom with a strong analytical and expert profile, where high editorial quality is our main asset. For a long time, this commitment to accuracy and expertise made us cautious about adopting non-traditional tools such as AI. However, with the growing role of AI in journalism and in society at large, we cannot ignore its impact.

As the role of AI in journalism and society continues to grow and can no longer be ignored, European Pravda is introducing internal rules that safeguard quality while also helping us work more productively, ensuring that our use of AI upholds the highest ethical standards, protects data privacy, and remains fully compliant with legal and journalistic norms.

We believe that transparency about our editorial policy on AI is an essential part of these standards, which is why we are making them publicly available.

These guidelines are a living, working document. They will be reviewed and updated whenever there are significant developments in AI technology or newsroom practices.

This incorporates the suggestion to mention ethics, legal compliance and regular review.

Our use of AI supports journalism. It does not replace human accountability. We aim to ensure responsible, transparent and fair use of AI tools that strengthen accuracy, protect sources and maintain public trust.

Our core principles

• Accuracy: All AI outputs must be verified by humans.
• Transparency: Readers will be informed about general AI use through a dedicated page on our website. Certain types of AI-generated content should be explicitly labelled.
• Data protection: Sensitive or restricted materials must never be uploaded to public AI tools.
• Accountability: Editors remain responsible for all published content, regardless of AI assistance.
• Ethical use. AI must not distort facts, mislead audiences, or undermine trust.
• Fairness: AI must not reinforce bias or discrimination in content selection or framing.
• Continuous Review: These guidelines will evolve alongside technology, law, and newsroom practices.

Practical rules

All possible uses of AI in the newsroom’s work are divided into three categories:

1. Green zone: AI use allowed and recommended.
2. Amber zone: AI use allowed under specific conditions.
3. Red zone: AI use is not permitted in editorial work.

Green zone tasks

• Spelling and grammar correction.
  Permitted in recommendation mode, with AI suggesting corrections. The actual corrections must be made by a human.

• Audio transcription.
  Our experience shows that this is a low-risk tool that improves efficiency, but it does not eliminate the need for editorial oversight. In particular, interview transcripts must be reviewed by the staff member who conducted the interview.

• Text translation.
  Translating completed news stories or article headlines into Russian using verified AI tools is a basic tool for creating the Russian-language version of the website and does not require additional labeling.
  Translating quotations from other languages is permitted but requires strict editorial oversight.

• Routine administrative or research tasks (brief text summaries, planning, schedule management).
  For internal use only. If used publicly, more thorough human review is required, and the task moves into the "amber zone".

• Generating YouTube thumbnails.
  The process must remain fully under human oversight; AI tools are used only to speed up the graphic-design work.

• Voice-over for videos on the YouTube channel.
  The voice-over script must be generated by a human; AI tools are used only to convert the text into speech.

• Drafting social media posts and headlines.
  These must be edited by a human before publication.

Red zone tasks

• News articles, analysis or opinions generated by AI.
  Prohibited under all circumstances. This rule also applies to freelance contributors, and violations will not be tolerated.

• Using AI for fact-checking or verifying the accuracy of claims.
  Prohibited due to the excessively high risk of errors and misinformation.

• Deepfakes or synthetic images, audio or video presented as factual.
  Strictly forbidden. For this reason, when generating YouTube thumbnails, it is important to avoid creating images that could be perceived as real photographs.

• Uploading strictly confidential materials or materials that could reveal sources or personal data to public AI tools.

Amber zone tasks

• Generating video or images.
  Allowed only if it does not mislead the audience and with the editor’s prior approval; the use of AI must be clearly disclosed to the audience.

• Data analysis or visualisation using AI.
  Human verification of findings is required, with a presumption that the AI may have made an error. Visualisations must be labelled as AI-generated.

• SEO optimisation, automated insights and complex summaries.
  Editor discussion is mandatory; the accuracy of the output must also be verified.

• Any task that does not clearly fall into the "Green" or "Red" zone.
  Editor discussion is mandatory. Log use of AI (date, tool, purpose, notes) if requested by editor.

Note: When carrying out "amber zone" tasks, human review for accuracy, tone and context of the generated content is mandatory.

Journalists are expected to share their findings and lessons learned during team meetings.

Closing statement

AI can strengthen our journalism if we use it consciously, critically, and transparently. Our goal is not to automate analysis, but to enhance human expertise through thoughtful, ethical, and collaborative adoption of new tools. By maintaining our editorial standards and openness to innovation, European Pravda will continue to combine credibility with efficiency, ensuring that technology serves our journalism, not the other way around.