#Morocco #Arabic - Morocco’s Ministry of Digital Transition and Administrative Reform (MTNRA) has released the first open-source AI tools built under its partnership with French AI company Mistral, designed to recognise and transcribe Moroccan Darija. The release includes a language identification classifier that distinguishes between Arabic dialects and Voxtral, a speech recognition model that transcribes Darija speech into text, including conversations that switch between Arabic, French and English. The tools support Morocco’s Maroc Digital 2030 strategy and AI Made in Morocco roadmap, led by Minister Amal El Fallah Seghrouchni.
SO WHAT? - Most large language models handle colloquial Arabic poorly, because the overwhelming proportion of Arabic text is written in Modern Standard Arabic (MSA), making training for dialects difficult. Releasing these tools as open source means any Moroccan developer, startup or government agency can build on them immediately, rather than struggle with solving the same dialect problem themselves. This is also a step forward for the government’s plans to build sovereign AI capabilities and involve more Moroccans in the digital economy.
KEY POINTS:
Morocco’s Ministry of Digital Transition and Administrative Reform (MTNRA) and Mistral AI have released the first two open-source AI tools from their strategic partnership: a language identification classifier and an automatic speech recognition model.
The classifier can recognise and distinguish between various Arabic dialects, including Moroccan Darija, while the speech recognition model, built on Mistral’s Voxtral technology, transcribes spoken Darija into text.
Voxtral can handle code-switching, meaning it can transcribe conversations that shift between Arabic, French and English mid-sentence, a common pattern in everyday Moroccan speech.
The tools are designed to help government agencies, startups, researchers, developers and businesses build AI applications that can understand and converse in Moroccan Arabic dialect.
Target applications include public services, education, media, customer support, document digitisation and multilingual data processing.
The release forms part of the AI Made in Morocco roadmap and the Maroc Digital 2030 strategy, aimed at developing sovereign AI tools tailored to Morocco’s linguistic and cultural context.
The tools complement a wider suite of AI solutions MTNRA develops and trains in-house to power its own AI marketplace and support public administrations.
The ministry said additional models and use cases are under development to strengthen national capabilities in generative AI, language processing, speech recognition and digital public services.
The release builds on a September 2025 memorandum of understanding between Morocco and Mistral.
ZOOM OUT - Minister Amal El Fallah Seghrouchni and Mistral AI co-founder and CEO Arthur Mensch signed an agreement in September 2025 focused on education, applied research and knowledge sharing in AI. The deal aimed to grow Moroccan startups, build AI tailored to local needs, with a strong emphasis on ethical, inclusive and responsible use and data protection. A year on, this week's Darija tools are the partnership's first concrete output.
[Written and edited with the assistance of AI]
Source: MTNRA
Read more about recent Morocco AI news:
Morocco to train 100,000 AI workers annually (Africa AI News)
Morocco Union calls for protection of outsourcing jobs (Middle East AI News)
Oracle expands Morocco footprint with Agadir R&D hub (Africa AI News)
Morocco signs deal for new YouCode Errachidia coding school (Africa AI News)
Morocco launches Rally AI Future Lab to accelerate innovation (Africa AI News)
there is any public repo in github to see the actual project