ICTworks

Here’s what I find maddening about the current state of AI governance in humanitarian organizations: we’ve got 80% of aid workers using AI tools daily while fewer than 25% of their organizations have any formal AI policies.

That’s organizational negligence disguised as prudence.

Too many humanitarian organizations are paralyzed by the complexity of AI governance, creating elaborate ethics committees and month-long approval processes that accomplish nothing except slowing down deployment while staff continue using ChatGPT and Claude without any guidance whatsoever.

Meanwhile, Charles Sturt University quietly released the SECURE framework that turns AI risk assessment into a simple six-question checklist.

Sign Up Now for more AI insights 

Governance Theater vs. Practical Solutions

Let’s be honest. Recent research from the Humanitarian Leadership Academy reveals that despite widespread individual adoption, only 8% of organizations report widespread AI integration and 64% provide little to no AI training for staff.

This creates exactly the wrong incentives.

Staff are using AI tools anyway. They just can’t be transparent about it or get institutional support when things go wrong. Organizations pretend they’re being “responsible” by having lengthy AI ethics discussions while their people are already deep in the AI adoption curve, making decisions with commercial tools that may not align with humanitarian principles.

I’ve lost count of how many humanitarian technology conferences I’ve attended where panels spend 45 minutes discussing the philosophical implications of AI bias while field staff in the audience are quietly wondering whether they can use Claude to translate beneficiary interviews without violating data protection policies.

Why AI Governance Frameworks Fail

The fundamental problem with most AI governance approaches is they’re designed by committees for committees. Look at the 53+ different AI ethics guidelines floating around the development sector—each one longer and more abstract than the last.

These frameworks typically fail because they:

1. Require expertise organizations don’t have: Most guidelines assume deep technical knowledge about AI systems that humanitarian organizations simply don’t possess.
2. Create approval bottlenecks: Complex review processes mean simple use cases get stuck for weeks while genuinely risky applications slip through with minimal scrutiny.
3. Focus on perfect solutions: Academic approaches that demand comprehensive impact assessments for every AI use case guarantee that nothing gets deployed.
4. Ignore operational reality: Beautiful principles that fall apart the moment you try to apply them in a refugee camp with intermittent internet and three overworked program officers.

SECURE Framework: Just Six Questions

SECURE framework takes a radically different approach. Instead of abstract principles, it provides six concrete yes/no questions that any humanitarian worker can answer:

1. Security credentials: Are you entering login details, passwords, or API keys?
2. Ethical use: Are you working with vulnerable populations or rare cultural knowledge?
3. Confidential information: Are you sharing data vital to organizational competitive position?
4. Use of personal information: Are you processing personally identifiable information?
5. Rights protection: Are you using copyrighted materials without permission?
6. Evaluation of outputs: Are you using AI output without human review for critical decisions?

If you answer “no” to all six questions, you’re cleared to proceed. If you hit “yes” on any category, you pause, try to mitigate the risk, and escalate to IT if necessary. That’s it.

No committees, no month-long reviews, no philosophical debates about the nature of algorithmic bias.

How SECURE Accelerates Deployment

This is where the framework gets clever. By clearly defining what requires approval, it dramatically speeds up everything that doesn’t.

Currently, most humanitarian organizations operate under blanket uncertainty. Staff assume they need permission for everything AI-related, so simple use cases get stuck in approval processes designed for complex ones.

The SECURE framework flips this dynamic.

Staff can confidently use AI for report writing, translation, data analysis of anonymized datasets, and content generation without waiting for organizational approval. This covers probably 80% of current humanitarian AI use cases.

Meanwhile, truly risky applications, like processing beneficiary personal data or making resource allocation decisions, get flagged for proper review. The framework channels organizational oversight energy where it’s needed instead of spreading it thin across every possible use case.

Why This Matters for ICT4D Practitioners

The humanitarian sector’s AI paradox of high individual adoption, yet low organizational support, is a classic implementation failure that we’ve seen repeatedly in ICT4D solutions. It’s the same pattern we saw with mobile money, social media adoption, and cloud computing integration.

Organizations that figure out practical governance frameworks first will have enormous advantages. They’ll be able to harness their staff’s existing AI skills, deploy solutions faster, and iterate based on real use cases rather than theoretical scenarios.

I predict that humanitarian organizations using frameworks like SECURE will be deploying AI solutions 3-6 months faster than competitors stuck in governance theater. In crisis response, that speed advantage translates directly to lives saved and resources optimized.