As international aid agencies adopt digital tools, using artificial intelligence to create and review deliverable documentation offers both benefits and difficulties. This guide outlines key ethical, contractual, and security considerations and highlights the benefits of responsible AI use.
Ethical Considerations
AI use in documentation must be consistent with the humanitarian principles of humanity, impartiality, neutrality, and independence. Since AI generates text based on existing data, it can reinforce or amplify biases. AI trained primarily on English-language sources from Western institutions may marginalise local voices or miss cultural context.
Transparency is a fundamental ethical requirement. Beneficiaries, donors, and partner organisations should be informed when AI has been used to draft or vet documentation. Accountability must remain with the agency and its staff; an AI tool cannot be held responsible for errors or misrepresentations.
Contractual Considerations
Many donor contracts specify that deliverables must be original work, and the use of AI-generated text could be interpreted as a breach of contract if not clearly communicated. Aid agencies should review their existing contracts and seek clarification where necessary.
Most AI platforms do not transfer ownership of generated content to users, and some providers retain broad rights to use, modify, or redistribute the text. This conflicts with the confidentiality required in donor-funded projects.
Security Risks of High-Risk Data Leaking
The most immediate and serious risk is the potential leakage of high-risk data. Aid agencies handle sensitive information such as beneficiary data, security assessments, financial records, and confidential communications. Most public AI tools retain and learn from user inputs.
Agencies should implement a "never type" checklist including: personally identifiable information, social security numbers, passport numbers, bank account details, and similar sensitive data.
Benefits of AI When Used Correctly
When used correctly, AI can significantly reduce the time spent on routine writing tasks. AI tools can also enhance consistency and quality, checking deliverables for compliance with donor formatting requirements and identifying gaps in logic or evidence.
Language translation and localisation are areas where AI offers considerable value, particularly for agencies working across multiple linguistic regions.
Recommendations for Responsible Use
Aid agencies should adopt a formal board-level policy defining acceptable tools, information security, staff training, and mechanisms for oversight. Procurement of AI tools should be centralised and rigorously vetted, with preference given to vendors committed to data privacy, transparency, and ethical AI development.
By using AI to support, not replace, human expertise, aid agencies can benefit from its capabilities without compromising their values or obligations.