This post is simply an informative piece and has no political bearing.

I recently came across an article stating that ‘un assistant pédagogique alimenté par l’intelligence artificielle’ (basically LLM/genAI) is being rolled into select classrooms. Brought it up with some friends, and all they heard was the local press coverage being positive. They were mostly of thought that this is a good solution and way forward considering the last 10-12 years in the educational sector.

After a lengthy conversation, I decided to make this post and give my opinion on the adoption of genAI/LLM/AI in education specifically in Mauritius.

Because we’re about to do this to an entire generation of Mauritian students with no publicly available local evidence that it works.

First, there are already other countries that have been testing AI in education. Let’s take a look at some related news articles:

Uh-oh. That’s definitely clickbait. We really need AI in education. Let’s instead see what the experts are saying:

Alright, there’s a lot and I barely scratched the surface researching this specific context. But what about our specific situation here in Mauritius?

The LLM in question

I was curious about mytGPT. On its website, it says ‘mytGPT is Mauritius’ first sovereign AI platform’.

First sovereign AI platform? Amazing.

However, I haven’t seen any technical report, whitepaper or anything related to the infrastructure, governance or the model being used. And I have so many questions.

I did contact its marketing team Dec 2025, and they were amazingly quick in setting up a Teams meeting. However, I had lots of technical questions and was told someone from the tech team would contact me.

Anyway, my questions I have a list of these but here are some of them:

  • Is it a locally developed LLM?
  • On what data was it trained?
  • Where can I see the weights or architecture?
  • Or is it an API wrap for ChatGPT, claude, etc.?
  • I haven’t found any reports publicly available regarding mytGPT and the data protection act.
  • The datacenter in question fully support this platform?
  • What exactly is meant by phonetic Creole? Is it Kreol Morisyen specifically or just phonetic creole?
  • What was the results of the pilot in Collège Maréchal in Rodrigues? Did students perform better/worse? Were teachers, psychologists and academia people consulted?
  • Can parents opt out of having their kids using this tool? What happens to the data collected by the students? Who owns it?

If someone has some answers to these, it would be much appreciated.

Final thoughts and concerns

What happens when students start doing their homework with AI and then have the LLM correct them instead of teachers? Would this adoption lead to cutting down funding in education or taking public financing from the academic world down the road? And what about the inevitable ‘why should we pay you more than the minimum wage since the LLM/genAI/AI tool is doing the heavy lifting’?

Anyway, for those saying this is a good tool and the LLM will explain its process or reasoning to students and teachers. LLMs don’t know why they give you an answer. They’re just really good at predicting what word should come next. When an LLM “explains” something, it’s simply generating more plausible-sounding text based on patterns it’s seen before. The real question is whether the explanations still provided are accurate and helpful.

Personally, I am against introducing LLM/genAI as a learning tool for kids. At that age, students need to be building critical thinking skills. Instead, empower the teachers and administrative staff with the tool to reduce their workload.

I feel there’s a FOMO and a rush to adopt this technology without consulting teachers or relevant parties directly involved or impacted. There should be another way to verify whether it actually helps students learn instead of directly introducing it in schools. Some countries that jumped in earlier than Mauritius are already rolling back their programs (just have a look at the links above). Mauritius doesn’t have to repeat their mistakes. We just have to ask some basic questions before we get too deep in instead of a ’trust me bro’ it works.

Although, I’d love to be proven wrong here. If you have thoughts on this, do reach out.