For the past 12 years, the Endangered Languages Project has been a trusted place to share digital resources to support Indigenous and endangered languages. People around the world have chosen to share information about their languages, cultures, and work through ELP, trusting that we will engage with their materials appropriately, respectfully, and transparently. 

Our relationships with language champions, community organizations, scholars, and allies are the core of our work. The trust our community places in us is at the heart of what we do.  

We are deeply committed to Indigenous data sovereignty, and to the rights of Indigenous and minoritized communities to decide whether, when, and how they will engage with language technologies.

That is why we have been deeply concerned, in recent months, to see false information circulating about ELP’s involvement in AI projects. Several posts and articles, all of which seemed to be AI-generated, have falsely claimed that ELP is developing large language models (LLMs) or other AI tools, using the language materials shared on our site. 

Using the language materials shared on ELP’s site to train LLMs would be profoundly unethical, and a gross violation of our values and relationships. We want to clarify ELP’s use, or non-use, of AI tools in general, and contextualize these decisions within our organizational values. 
 

• First and foremost, we do not engage in language data mining using LLMs or other AI tools. We do not permit materials hosted on the ELP site to be used in the training of LLMs or other AI models.
• We work with human writers and editors - we don’t use ChatGPT or other generative AI tools to create our learning resources, social media content, articles or blog posts. We invest significant collective time, thought, and emotion in writing the text you find on the ELP website, social media accounts, and learning materials. We value foregrounding a variety of human voices, writing styles, knowledges, and creative visions. Working together with human writers and editors creates possibilities for one of our core values of relationship-building. 
• We work with human artists - we don’t use generative AI image or music/sound generation tools for our materials. We do not knowingly publish AI-generated images. We value experiencing and sharing a variety of human art styles, perspectives, artistic traditions, and creative visions.
• We work with human translators and interpreters - we don’t rely on machine translation tools without human supervision. We sometimes use transcription tools for materials in dominant languages, or use machine translation tools to quickly check the meaning of a piece of text, but always check any machine translations with a human. We value and respect the human knowledge of context, tone, style, and cultural nuance that goes into translation work.
• We work with human content creators - we aim to support and uplift creators who put time, effort, and creativity into conducting interviews, recording audio and video materials, creating social media posts, etc. We value elevating the perspectives and talents of humans around the world, and creating space for people to learn from one another.
• We work with human researchers - the information in ELP’s Catalogue of Endangered Languageshas been painstakingly collected, organized, and reviewed by a team of human researchers over many years. In all of our research work, we do not use ChatGPT or other LLMs. We do our own thinking, reading, editing, and writing.

We suggest that anyone interested in the development of AI technologies for language revitalization, in alignment with the values of Indigenous data sovereignty, may wish to learn more about the work of organizations like Indigenous AI, Te Hiku Media, First Languages AI Reality (FLAIR), Te Mana Raraunga,Maiam nayri Wingara,Masakhane, the Global Indigenous Data Alliance, and others. 

As the technologies described broadly as “AI” continue to grow and change, and more ethical tools are developed, our use of AI tools may change. However, any choices we make about AI will remain informed by our organizational values: integrity, respect, support, optimism, collaboration, relationship, sharing and accessibility, and informed action. And, in keeping with our values, we will continue to communicate transparently about our decisions and actions regarding AI tools.

We hope this statement clarifies how ELP engages with AI, and we welcome any questions or comments you might have: feedback@endangeredlanguages.com