As advancements in artificial intelligence continue and access to AI-powered tools becomes more widespread, the humanitarian aid sector is being faced with both new opportunities to embrace innovative approaches towards the enhancement and expansion of access to aid while also confronting significant challenges in ensuring responsible AI adoption and implementation.

Building on the decade-long partnership between RSN and Refugee Rights Turkey (RRT) dedicated to improving the legal protection of refugees in Turkey, this report reflects on RSN’s exploration of Large Language Models (LLMs) to power chatbots aimed at enhancing access to critical legal information for asylum-seekers.

Reflecting on the experience of building this prototype tool, the report examines core concerns with AI implementation in humanitarian contexts such as factual inaccuracies, algorithmic bias, and privacy issues. The report also includes preliminary findings from the prototype chatbot, underscoring both the potential benefits and significant investments needed to responsible use AI technologies in sensitive settings.

The report concludes with a set of actionable recommendations for humanitarian actors and funding organizations, emphasizing the importance of community involvement, sufficient ethical considerations, and robust AI application monitoring.