Safe Use of AI - Guidelines for NGO Teams

AI can help NGOs work faster, smarter, and more efficiently. But it also comes with risks - from spreading misinformation to violating privacy or unintentionally harming the communities we serve. Safe use means balancing innovation with responsibility, ethics, and transparency.

This page offers practical steps to ensure your AI use aligns with your mission, values, and legal obligations.

⁠

Principles for Safe AI Use

Principle

What it Means

Example

Human Oversight

People remain in control; AI assists, not replaces.

Staff reviews AI-generated reports before sharing internally or externally.

Transparency

Be clear when and how AI is used.

Note: “This report contains AI-assisted analysis.”

Privacy

Protect personal/sensitive data from misuse.

Use anonymized data. Never upload any data set with personal identifier information to AI tools (such as name, phone no., email id, address, aadhaar or other govt id numbers, etc).

Bias Awareness

AI can reflect and amplify social biases.

Test chatbot responses for discriminatory language before widely sharing

Accountability

Assign responsibility for AI use.

Ideally, if possibly, internally, orgs could assign one team member (be it tech or non tech team) to ensure to test AI tools before widely using in the org.

⁠

Key Risks & How to Mitigate

Risk

Examples

How to Mitigate

Data Privacy

Storing sensitive beneficiary data in public AI tools.

Check tool privacy policies; anonymize data; store securely.

Misinformation

AI outputs that sound right but are wrong.

Fact-check against trusted sources.

Bias

AI disadvantaging certain groups.

Test outputs for fairness; involve diverse reviewers.

Legal Issues

Verify sources; review agreements before use.

Reputation Damage

Perception of replacing people with AI.

Communicate AI’s supportive role; involve communities.

⁠

Do’s and Don’ts

✅ Do

Keep humans in the loop for all critical decisions.

Pilot AI internally before external use.

❌ Don’t

Share confidential data with public AI tools.

Assume outputs are correct without review

Use AI to make decisions affecting rights/welfare without human oversight.

⁠

Safe AI Use Workflow

Define the Purpose - What problem will AI solve?

Assess Risks - Privacy, bias, misinformation.

Select the Tool - Prefer strong privacy & transparency policies.

Test - Small internal pilot; review accuracy & bias.

Implement with Oversight - Human review checkpoints.

Monitor & Improve - Track performance, feedback, and incidents.

⁠

Critical NGO Contexts

Vulnerable Communities - Extra care with consent and anonymity.

Advocacy & Policy - Fact-check every claim in AI-generated briefs.

Fundraising - Avoid generic/misleading AI narratives; respect donor data.

Education - Ensure local relevance in AI-generated materials.

⁠

Quick AI Safety Checklist

Purpose of AI use is clear.

Privacy/consent requirements met.

Risks identified & mitigated.

Human reviewer assigned.

Outputs fact-checked.

Ongoing monitoring in place.

⁠

Red Flags - Stop Using AI Immediately If…

AI outputs contain sensitive personal details that were not supposed to be shared.

The tool generates harmful, discriminatory, or offensive content.

You cannot explain how a critical AI decision was made.

You spot serious factual errors that could mislead beneficiaries, partners, or donors.

There is a data breach or unapproved sharing of information with third parties.

The AI’s output undermines trust with the community or key stakeholders.

⁠

Capacity & Culture

Train staff regularly on AI basics and ethics.

Appoint an AI focal point for oversight.

Share internal case studies of AI successes and failures.

⁠

Resources

⁠

UNESCO AI Ethics Recommendation⁠

⁠