OpenAI's AI finds 948 flaws in Linux — and 81 in Firefox
The new GPT-6 Cyber model scanned the Linux kernel and the Firefox browser, finding hundreds of previously unknown flaws — and developers are now fixing them.
Read the story →AI news
AI news in Kiswahili
News, knowledge and practical uses of artificial intelligence, explained in simple language.
Who are you? We'll show what's relevant to you.
Or browse by interest
This week's news
All news →
IFC invests in Kifiya: the AI helping farmers get loans without collateral
The International Finance Corporation has invested in Ethiopian fintech Kifiya, whose AI assesses borrowers with no credit history — and the company is now looking to expand into Tanzania, Kenya, Uganda and Rwanda.
South Korea lends $170 million to build an AI training centre in Tanzania
The agreement signed in Dar es Salaam on Wednesday will build the first centre of its kind in East Africa — teaching AI, robotics, data analytics and the Internet of Things.
Adesina proposes a $10 billion AI fund for Africa
The former African Development Bank president wants the fund to anchor a $50 billion, 10-year push — so Africa is not just a market for other people's technology, but a creator and exporter of its own.
Google launches Skilling Africa: learn AI for free in Swahili, even offline
Google's new hub, launched on November 2, gathers more than 15 of its training programmes in one place — free, in Swahili, with lessons you can download and study without internet.
DeepSeek open-sources a free AI coding assistant that works on its own
China's DeepSeek published on October 31 an open-source experimental system that listens on GitHub and does programming work — but it warns plainly: it is not a finished product.
East African ministers approve regional AI strategy and university AI module
Ministers of the East African Community have approved a shared AI strategy and a university AI module at the Kampala meeting that ended on October 31 — but implementation still has no deadline or new funding.
AI giants agree on one signal: websites can now block their content from training AI
OpenAI, Google, Microsoft and Anthropic announced on October 30 that they will honour one shared signal in robots.txt — instead of separate rules for each company — so a website owner can decide whether their content is used to train AI.
Mistral gives away a trillion-parameter AI model: anyone can download and run it themselves
French company Mistral has released the weights of its new Large 4 model — built to find and fix security flaws — so organisations can run it on their own servers instead of depending on another company.
Google launches Gemini 4 Carbon: a new AI to defend software against hackers
Google DeepMind's new model says it finds security flaws 88% of the time and patches them itself — but for now only cyber defenders get access, not the public.
Microsoft offers exit deals to 60,000 staff as it pours billions into AI
The world's biggest software company is offering more than a quarter of its workforce a voluntary exit — while planning to spend $175 billion building AI infrastructure this year.
Anthropic goes public: Claude's maker reveals $16.3B revenue and an $80.7B loss
The US AI company has filed to list on the Nasdaq stock market — but most of its huge loss is a paper accounting charge, not cash lost.
OpenAI cuts prices by up to 90% to lure Anthropic's biggest customers
OpenAI is offering steep discounts to Anthropic's biggest customers — and some have already switched, according to a report by The Information.
Nothing for this group yet. We're writing it. Tell us what you want to know.
Tools we've tested
We tested: Gemini in Kiswahili
Strong on translation and images, with one real advantage: it can search the web.
We tested: ChatGPT in Kiswahili
Its Kiswahili is good for everyday writing, but it slips on present-day Tanzanian specifics.
Learn step by step
Small business: five ways to use AI at no cost
You do not need to change your business. Start with one task that tires you out every week.
How to ask AI well: a guide to writing prompts
A bad answer usually comes from a bad question. These are the four parts of a prompt that works.
Start using AI today: five steps on your phone
You do not need a computer or money. You need a phone with internet, ten minutes, and one real question.
Prompts from the community
Share yours →From Pashavo · used 1
Write a short WhatsApp Status description for this product: [product, price, town]. Three lines, everyday Kiswahili, no exaggeration.
From Pashavo · used 0
Summarise this text in two paragraphs of simple Kiswahili, then list the three key points: [paste text]
From Pashavo · used 1
Write an application letter for a [job] role at [company]. Twelve lines, official Kiswahili, using this experience: [your experience].
From Pashavo · used 0
Explain [hard concept] as if to someone who has not been to university. Use an everyday Tanzanian example. No more than 150 words.
Get the news every Monday, on WhatsApp
One message a week. No adverts, no charges, no middleman. Reply STOP any time.
What do you want to know?
Ask anything about AI. Many questions become stories here.