Last updated: May 21, 2024
Generative AI Usage Policy
At Triba, we are committed to positively contributing to environmental, social, and economic well-being while delivering value to our clients and stakeholders.
This is why we adopt best practices for responsible generative AI based on the recommendations of the Canadian government.
To maintain the trust of our stakeholders and ensure responsible use of generative AI tools, we adopt the “READY” principles developed by the Treasury Board of Canada Secretariat (1):
1 – Relevant
Ensure that the use of generative AI tools meets the needs of users and the organization; consider environmental impacts when choosing to use a tool; determine the appropriate tools for the task; AI tools are not the best choice in all situations.
2 – Responsible
Take responsibility for the content generated by these tools and the impacts of their use, including ensuring that the generated content is accurate, legal, ethical, and complies with terms of use; establish monitoring mechanisms.
3 – Fair
Ensure that the content of these tools does not contain or amplify biases and respects human rights, accessibility, and equity obligations.
4 – Transparent
Identify content produced using generative AI (see Appendix 2); inform users that they are interacting with an AI tool; provide information about institutional policies, appropriate use, training data, and model when deploying these tools; document decisions and be able to provide explanations if tools are used to support decision-making.
5 – Informed
Learn about the strengths, limitations, and responsible use of tools; learn how to create effective guiding messages and identify potential weaknesses in results.
6 – Secure
Ensure that the infrastructure and tools are suited to the security classification of information and that privacy and personal information are protected; assess and manage cybersecurity and reliability risks when deploying a system.
Appendix I - Examples of Appropriate Use Cases
Here are some examples of use cases for the use of generative AI at Triba. The purpose here is not to establish an exhaustive list but rather to provide concrete cases so that you can improve your understanding of best practices and exercise judgment when using generative AI.
Image Generation to Illustrate Articles
The use of tools such as Meta AI, DALL-E, or RunwayML to generate images is appropriate, as long as the iterations remain limited and the images are clearly identified as coming from generative AI (see Appendix 2 for best identification practices). Since generative AI consumes a significant energy source, it is important to always ask the following question: can I support my message with an existing image? Is the energy cost of generating this image proportional to its added value?
Translation
The translation of content for a blog or article can be done using generative AI tools. However, it remains important not to copy any information that is not public knowledge. For example, excerpts from an organizational diagnostic report for a client or excerpts from our methodology that are considered our intellectual property should not be found in generative AI tools.
Document Review to Improve Syntax or Correct Spelling Errors
The same guidelines apply as for translation.
Idea Generation to Support Brainstorming
Recent studies show that generative AI can hinder creativity and innovation(2). We recommend using AI as the last step in the brainstorming process. Our best practices are to brainstorm without technology and then use AI to generate new ideas or build on ideas already generated by our team.
Content Creation, with Restrictions
The draft framework for writing an article can be done using generative AI, but the article writing and information validation must be done by a human. Our mission is to create value for our stakeholders, and this value is diminished if we use and publish content generated solely by AI.
Appendix 2 - Identification of AI-Generated Content
There are currently no standards for identifying AI-generated content. These guidelines are therefore subject to change based on new developments.
The priority is to clearly identify whether and how AI has been used to generate content.
