Hi everyone,

I’m Munenyashaishe (Ishe) Hove, an AI governance and safety researcher from Zimbabwe.

My background is a bit non-linear. I started in accounting, finance, and audit, then moved into data science, AI teaching, digital transformation, responsible AI, and eventually AI governance and AI safety. I am currently pursuing a PhD in Information Systems and Technology at the University of KwaZulu-Natal.

The problem I keep coming back to is this:

Many organizations are being encouraged to adopt AI, evaluate AI, monitor AI, audit AI, and manage AI risks. But in many resource-constrained contexts, organizations may not even know where AI is being used, what risks are emerging, who is responsible for monitoring those risks, or what should happen when a risk signal appears.

My PhD focuses on operationalizing AI safety governance in resource-constrained African organizational contexts. I am looking at two linked gaps:

1. The AI Risk Visibility Gap

This is where organizations lack the tools, awareness, internal controls, inventories, reporting channels, and monitoring systems needed to make AI-related risks visible in the first place.

2. The Evaluation-Governance Gap

This is where AI risk evidence exists, for example from evaluations, audits, incidents, benchmarks, red-teaming, monitoring, or user complaints, but is not translated into structured, auditable, and defensible governance decisions.

In simpler terms, I am interested in what happens after a risk is noticed.

Who investigates it?
Who classifies it?
Who escalates it?
Who decides whether to monitor, pause, restrict, audit, redesign, or stop an AI system?
How is that decision documented?
What makes the process accountable?

This interest comes partly from my audit background. Before moving into AI, I worked in audit and spent time thinking about controls, evidence, documentation, risk, and accountability. When I later moved into data science and AI governance, I started seeing similar questions appear in a new form.

I have also been building my AI safety understanding through programs such as BlueDot Impact’s Technical AI Safety course, BlueDot AGI Strategy, AI Safety South Africa, ML4Good, and EA-related programs. One technical project I worked on involved fine-tuning TinyLlama models on synthetic false information and examining how the model’s behavior changed. That project pushed me to think more seriously about evaluations, risk signals, and what organizations should do with evidence about model behavior.

Right now, I am trying to figure out where I can contribute most effectively within the AI safety ecosystem.

My current hypothesis is that my strongest fit may be technical AI governance or operational AI safety governance, especially work that connects evaluations, incident reporting, monitoring, audits, and institutional decision-making.

I would especially value feedback on:

• Whether the AI Risk Visibility Gap and Evaluation-Governance Gap seem like useful framings  
• Which organizations, researchers, or projects are working on related questions  
• Whether this direction seems more useful for AI governance, AI evaluations, AI assurance, policy implementation, or field-building  
• What outputs would make this work more useful to labs, governance organizations, funders, or policymakers  
• How someone from an African and interdisciplinary background can build a stronger path into AI safety and governance work  

I am also interested in connecting with others thinking about AI safety and governance from African, Black diaspora, Global South, or resource-constrained institutional contexts.

If any of this overlaps with your work, I would be very happy to connect.