InfoLight.UA recognizes the importance and potential of artificial intelligence (AI) technologies to improve our fact-checking, data analysis and content creation. This policy sets out the principles and rules for the use of AI in our organization.
The purpose of this policy is to ensure the ethical, responsible, and efficient use of AI technologies at InfoLight.UA, while maintaining our commitment to accuracy, objectivity, and transparency.
Ethical principles
InfoLight.UA uses AI systems in accordance with the core values of journalistic ethics, including truthfulness, accuracy, impartiality, independence, non-discrimination, accountability, inclusiveness, respect for privacy and confidentiality of sources.
We do not tolerate linguistic, political, and other biases and stereotypes in the content generated by AI.
The main principles of using AI
- Responsible editorial decision
- Legality
- Regular risk assessment
- Transparency and clarity
- Audience awareness
- Privacy and data protection
- Professional human control
Areas of AI application in InfoLight.UA
- Sounding of news
- Deciphering texts (video transcription)
- Formation of headings and SEO blocks
- Partial rewriting of news texts (not original research!)
- Generate images
- Creating entertainment content
- Analyze data and identify trends
- Translation assistance
Our technical partner is Osavul, an AI-based malicious narrative tracking system!
Rules for using AI
Mandatory editorial review: All AI-generated content is subject to editorial review. InfoLight.UA considers such content as unverified source material. The final professional control and responsibility for the content created by AI in whole or in part lies with a human.
Content labeling: Images generated by artificial intelligence are necessarily labeled.
News voiceover: When voicing content, InfoLight.UA does not use a synthesized voice that would try to duplicate the voice of a real person, except for voices created based on the voices of our team members.
Text transcription: AI is used to automate the transcription of texts from videos, but editors must proofread them for literacy and relevance to what the speaker said.
Formation of headlines and SEO blocks: Editors can use AI suggestions, but priority is given to their own creativity and originality.
Creating entertainment content: If you create tests or similar content that may contain an interpretation of psychological problems, add a disclaimer about the need to consult a specialist.
Limitations and liability
- InfoLight.UA recognizes that AI has limitations and may miss or distort important information.
- InfoLight.UA remains completely human-centered. A human evaluates the effectiveness of AI tools and makes decisions on the publication of any product created or processed by AI.
- InfoLight.UA reserves all exclusive property copyrights to objects created or processed by AI.
Training and competence
All employees who use AI tools are trained on the capabilities and limitations of these technologies.
Monitoring and evaluation
InfoLight.UA regularly reviews and evaluates the use of AI to ensure compliance with this policy and general standards.
Compliance with EU legislation
InfoLight.UA undertakes to comply with the requirements of of the Regulation of the European Parliament and of the Council on harmonized rules for artificial intelligence (Artificial Intelligence Law). In particular, we:
- We conduct risk assessments for all AI systems we develop or use.
- Ensure compliance with the requirements for high-risk AI systems, if any.
- We provide the necessary documentation and transparency about our AI systems.
- We ensure proper human oversight of AI systems.
- We inform users about the interaction with AI systems.
- We cooperate with the relevant regulatory authorities.
We regularly review our policies and practices to ensure ongoing compliance with EU artificial intelligence legislation.
Policy update
This policy is regularly reviewed and updated in accordance with the development of AI technologies and the needs of the organization.
Leave a Reply