Global Canopy statement on AI usage for communications

Transparency and accountability are at the heart of Global Canopy. We are committed to rigour and honest inquiry as we work to promote informed, evidence-based action for nature, forests and people.

As a data-driven organisation, we recognise that advancements in AI can offer exciting new possibilities and efficiencies. At the same time, lack of care in how it is used can reduce the quality of work, proliferate misinformation and undermine trust.

This guidance aims to set clear expectations on when and how we may harness generative AI in the process of developing our published visual, text, audio and other communications content. Anyone involved in creating content for Global Canopy (whether they are staff, contractors or collaborators) must adhere to these guidelines.

Overarching principles

• We affirm that human creativity, reasoning and attention are at the core of all our work. We can judiciously use AI to enhance or support – but never replace – these qualities. 
• We take care to ensure that fictive scenarios are not represented as fact.  
• We recognise that AI (particularly Large Language Models) cannot reliably achieve accurate outputs. AI use must always be accompanied by human fact-checking and critical assessment.
• We know that AI output can be affected by bias and consciously seek to account for this possibility. 

Guidelines for specific kinds of content

Images, videos

• Global Canopy will avoid the use of photorealistic AI-generated images or videos in published content, unless there are fantastical or abstracted elements in the visual that make it unambiguous that an imaginary item is depicted (e.g. an armchair floating amid planets in outer space). Any such use will be disclosed.
• In particular, Global Canopy will not use AI-generated images or videos of natural ecosystems, agricultural production or other landscapes. Our work is rooted in a recognition that each geographical place is unique and should not be treated as interchangeable. 
• Global Canopy may use AI to generate initial drafts of abstract graphics, but their final use will always be subject to human review. We may use AI to carry out minor edits (e.g. adjust lighting or contrast) to enhance the clarity of an image or video, provided they do not alter the original meaning or context of the visual, and any use is reviewed by a human.

Text

• Global Canopy will never publish text that is wholly authored by AI tools, unless there is a clear reason for doing so (e.g. demonstrating AI tools for the purpose of analysis). If so, both use and reason will be disclosed. 
• AI may be used to inspire a writer during the initial storyboarding or  brainstorming phase on a piece of work, help structure content, or support research – such as suggesting further reading or finding sources of information – and correct grammatical errors. We recognise that AI cannot and should never replace human judgment or creativity, and thorough human review is essential. In particular, Global Canopy writers and researchers will not rely on AI tools to summarise sources without further fact-checking, due to the risk of inaccuracies, hallucinations and missing key highlights in the text. 
• Using AI to generate subtitles from recorded audio or visual can be a good starting point for creating transcripts. Global Canopy will only publish these with an additional layer of human oversight, although we may make third-party subtitling tools available for some of our content where the use of these tools and their limitations are clear and widely understood (e.g. automated captions on YouTube videos). 

Audio

• Music composed by a human is always preferred over AI-generated ones to support creativity and acknowledge creators’ contributions.
• We will not use AI voice clones impersonating an individual’s voice. If AI-generated voiceovers are used, they will be accompanied by disclosure. 

Translations

• Automated AI translations can be useful in the initial stage of research – helping to translate articles or reports to further understanding. 
• AI translations should not be used to create content for publication either internally or externally without thorough checking by a fluent speaker before use.
• Wherever possible, Global Canopy will use professional human translators for reports and insights, to ensure quality, accuracy and cultural sensitivity. Some communications may be translated by fluent speakers within the team. 

Privacy concerns

• Confidential and sensitive information will never be shared with third-party AI tools, as it may be leaked or used inappropriately. 

Third party communications

• Where other individuals or organisations publish AI-generated content which reproduces Global Canopy’s material without appropriate attribution, we will consider this plagiarism and may seek redress.