Skip to main content
AI POLICY

Summary

This policy applies to everyone at The Green Line, from journalists to non-journalists, serving as a foundation for developing ethical AI practices. Given AI’s evolving nature, these guidelines are intended to be dynamic and adaptable, revisited as technology advances.

The Green Line is committed to producing 100% AI-free journalism. We do not use AI in any of our public-facing editorial content. We do, however, use AI tools behind the scenes for business development and operations, which help build our capacity as a small, resource-constrained newsroom.

As AI becomes more integrated into our work, understanding its ethical implications is essential. Generative AI — a subset of AI that creates text, images, audio, or video — has begun influencing Canadian journalism, including approaches to content creation and audience engagement. Ethical frameworks are crucial for balancing AI’s benefits with potential risks, particularly in maintaining trust with our audience.

This policy defines key terms, references existing ethical principles, addresses AI-specific scenarios, and suggests methods to apply AI ethically to further The Green Line’s editorial and business goals. We treat AI as a tool that complements, rather than replaces, journalistic judgment. AI assists us in research and automation of repetitive tasks (e.g., transcriptions). This supports our journalists to focus more on in-depth analysis, fact-checking and nuanced reporting that requires human insight.

We are particularly vigilant about risks associated with deep fakes and other AI-generated misinformation, which could erode trust in journalism. To combat this, we emphasize transparency in our AI use and prioritize real-life interactions.

Key definitions

  • AI (Artificial Intelligence): Refers to systems that mimic human intelligence to perform tasks, such as data analysis, pattern recognition, and language processing.
  • Generative AI: AI capable of producing new content based on patterns from vast datasets. Examples include chatbots, image generators, and voice mimicking tools.
  • Transparency: Being clear about where and how AI is used in journalistic processes.
  • Privacy: Protecting the privacy of sources and subjects when using AI-driven tools.
  • Security: Ensuring data handled by AI is securely managed and does not lead to breaches or misuse.
  • Humanity: Preserving human oversight and judgment in AI-driven processes to uphold the integrity and empathy central to journalism.

Key considerations

  1. Principles over tools and guidelines: While guidelines are useful, it is more important for us to first define our core principles regarding AI use. These principles prioritize human accountability, transparency and the integration of ethical standards into the technology itself. Before adopting AI tools, we clarify our values — whether AI is intended to increase efficiency, expand audiences or improve editorial quality, for example — and consider the responsibilities that accompany these decisions.
  2. Guidelines are fluid: AI is rapidly changing. We regularly assess and update ethical guidelines  to remain relevant, and we prioritize values like human accountability when interacting with AI tools.
  3. Human in the loop: While AI can streamline many processes, we keep a human in the loop for quality control and ethical oversight.
  4. Room for failure: Experimenting with AI is encouraged, and we want our journalists to feel comfortable discussing their AI use internally. Overcoming the stigma around AI use is crucial, so staff should disclose AI use to their colleagues, not just our audience.
  5. Disclosing AI use: The level of disclosure will depend on the context. For example, AI in research does not need the same transparency as AI-generated audience-facing content. However, transparency is key to maintaining trust.
  6. Training for journalists: We get training on AI literacy. Our staff needs to understand the capabilities and limitations of AI tools, and learn how to use them responsibly.
  7. Environmental impact: We factor environmental impact into decisions about which AI tools we adopt, particularly when considering large proprietary models that rely heavily on cloud computing. We are open to exploring open-source and locally hosted models; these options are often more energy-efficient, offer stronger privacy protections and are easier to operate sustainably. Smaller models can typically run on existing newsroom hardware, reducing reliance on energy-intensive data centres and high compute power.

Existing journalism ethics principles applied to AI

  • Accuracy: AI can introduce errors or bias, which require rigorous fact-checking. Our staff must question the results of AI by spot-checking and verifying information.
  • Fairness: AI systems often rely on datasets that may reflect societal biases. Our staff must work to prevent AI from perpetuating inequities, and ensure tools do not disproportionately affect marginalized communities.
  • Independence: Our staff must remain critical of AI tools, especially when platforms themselves have ethical pitfalls. These include where the training data comes from (often proprietary, non-transparent sources) and potential conflicts between profit motives and public interest.
  • Transparency: We will label the use of AI in our public-facing processes to maintain audience trust.

AI-specific scenarios

  • AI-Generated content: Generative AI tools like chatbots or text generators can create summaries, transcriptions or entire articles, but what level of error is acceptable? Our staff will define error tolerance internally and ensure human oversight in these processes. We will also tackle deepfakes, which could manipulate content in ways that harm journalistic integrity, especially during sensitive periods like elections.
  • Labelling AI use: AI use should be transparent. This helps the audience understand how and why AI was used.
  • Bias in AI models: Our staff will consider the biases in AI tools, and decide which platforms align with our values. Tools that rely on biased datasets can skew the information we produce, exacerbating issues of fairness and representation.
  • Data security and privacy: AI can hoover vast amounts of information, and our staff will ensure that sensitive data — whether from sources or audiences — is kept safe. This requires understanding the privacy implications of AI tools and adhering to strict security protocols.
  • Business development: AI helps us create targeted subscription offers and personal data-driven marketing campaigns. We must balance revenue generation with public trust, ensuring our AI practices align with our core values.