By Nelly M. Nyangorora
Public consultations on Kenya’s draft Artificial Intelligence and Emerging. Technologies Policy end this week.
Before us is a document that many experts have described as impressive. It envisions sovereign data centres, green compute zones and a National AI Council to regulate frontier technologies.
It speaks the language of risk-based governance, digital sovereignty and ethical innovation.
But as media professionals who have spent decades amplifying voices that rarely reach the national frequency, we at the International Association of Women in Radio and Television (IAWRT) submitted our reflections with one question in mind: who gets to speak in Kenya’s AI future, and in what language?
The draft policy emphasises domestic compute infrastructure, sovereign cloud arrangements and local data hosting, with the stated aim of building systems that serve Kenyans. Yet on language accessibility, it adopts a non-committal tone. It proposes to “promote” the inclusion of local languages, equitable access and participatory design. In policy and legislative grammar, “promote” is the diplomatic cousin of “hopefully”: it signals intention without obligation.
In a country where English and Kiswahili dominate official discourse while more than 40 indigenous languages carry the daily weight of commerce, culture and community aspiration, “promote” without enforcement becomes another word for exclusion.
The Constitution of Kenya recognises indigenous languages and Kenya Sign Language as part of our national heritage.
Broadcasters, especially on radio, have long understood that reaching rural audiences requires speaking in Ekegusii, Dholuo, Borana or Kikuyu, not just standard Kiswahili. The AI policy that will govern automated public services, diagnostic tools and educational platforms must treat language as a right to be protected, not a feature to be added later.
If a farmer in Bungoma cannot query an AI-enabled agricultural extension service in her mother tongue, the algorithm is not serving her. If a deaf or hard-of-hearing person cannot access government services because a health chatbot excludes Kenya Sign Language, the system is not merely inconvenient; it is discriminatory.
If county governments are expected to deploy AI tools but are given no mandate to align them with the linguistic realities of their constituents, devolution becomes a mirage.
Globally, jurisdictions embracing AI governance are moving from voluntary guidance to mandatory multilingualism. They are treating language as a foundational design requirement.
According to India’s Press Information Bureau, Bhashini, a public digital infrastructure launched under the National Language Translation Mission, hosts more than 350 AI language models covering 22 scheduled languages, while the Project Vaani initiative is extending coverage to 59 languages and dialects through open datasets and open APIs. There, multilingualism is the default, enabling speech-to-text, translation and conversational AI in languages such as Tamil, Bengali and Punjabi. It demonstrates that low-resource languages can be served at scale when the state treats them as strategic infrastructure.
The draft policy is also silent on whose knowledge qualifies as data. Indigenous oral histories, community-held linguistic archives and culturally specific ways of knowing do not always fit neatly into conventional data protection frameworks designed around individual consent.
When a language spoken by a few thousand elders is extracted to train a commercial AI model without communal consent or benefit sharing, the practice is not ethical. Kenya needs safeguards that recognise indigenous language data as communal heritage, governed by free, prior and informed consent, not treated as raw input for model training.
We, as women in media, also ask: who is in the room when AI governance decisions are made? The proposed National AI Council includes seats for industry, academia and generic civil society representation. But technologies that shape information access, automated hiring, credit scoring and content moderation require voices that understand how these systems can disadvantage women, persons with disabilities and minority communities.
AI learns from biased data and is deployed by institutions with blind spots. The draft policy must not replicate existing inequalities. Women must be deliberately included at the AI decision-making table, and gender-disaggregated impact assessments must be required.
As the Ministry of Information, Communications and the Digital Economy moves from consultation to drafting, we urge it to consider the fundamentals captured in IAWRT’s submissions: mandate, rather than merely encourage, multilingual accessibility in all publicly funded AI systems; set clear enforcement timelines; recognise Kenya Sign Language in public-facing digital services; empower counties to align AI deployment with local linguistic realities; protect communal data sovereignty; make indigenous language communities stewards of their linguistic heritage; and diversify the National AI Council to include linguists, PWD representatives, community media practitioners and gender experts.
Kenya is viewed as a regional leader in digital innovation. Africa is looking to us to show that we can build an AI governance model rooted in our context, not a copy-and-paste version of Brussels or Silicon Valley.
We need AI systems that a pastoralist in Isiolo can use to access climate information in their language; that a deaf student in Nyeri can use to learn through AI-assisted KSL resources; and that a rural broadcaster in Shinyalu can use without surrendering her community’s voice to proprietary algorithms.
At IAWRT Kenya, we will continue to monitor, advocate and amplify, because meaningful connectivity is about being heard.
Nelly Moraa Nyangorora is an independent editor, journalist, and Ethics and Responsible AI (ERAI) Fellow with more than 28 years of experience in journalism, communications, and public policy.
She is a former Chief Parliamentary Editor at the Kenya Broadcasting Corporation (KBC), where she pioneered live commentary of parliamentary proceedings under Kenya’s Parliamentary Broadcast Unit, building extensive expertise in legislative and policy reporting.
She holds a postgraduate qualification in Reporting Climate Change from Università Telematica Internazionale UNINETTUNO, Italy, and writes on the intersection of artificial intelligence, climate, health, governance, and digital rights.