What if two phones could share what matters with no internet, no Wi-Fi, no hotspot and no Bluetooth? Set all four aside, and ask what the devices can still do on their own. That is the question behind ongoing research at SISU AI.
The question
SISU AI is researching how devices can share data and communicate without the internet, Wi-Fi, a hotspot or Bluetooth, using edge technology alone. By edge technology, we mean computing that happens on the devices themselves rather than in a distant data center.
Those exclusions define the research. Wi-Fi, hotspots and Bluetooth are what people usually fall back on when the internet is out of reach. Ruling them out makes the problem harder and the question sharper: with none of them available, what can the intelligence on a device do to help it share what matters?
The work is early and ongoing. This post sets out the questions we are asking.
Why it matters
Much of today’s software assumes a working connection. For a large share of the world, that assumption doesn’t hold. The International Telecommunication Union estimated that 2.2 billion people worldwide remained offline in 2025, and that 58% of people in rural areas were online, compared with 85% in urban areas.
Some of that gap is coverage, and some of it is cost. Many people live within reach of a mobile broadband signal but don’t use mobile internet, often because of the cost of data and devices.
Even where the network is there, it can fail. A power cut or a storm can take it down, in a city as well as a village. Picture a district after a flood: phones still hold some charge, people still need to share what they know, and the network is gone.
Help that depends on a signal ends where the signal ends.
What we are exploring
We have more questions than answers at this stage. These are the ones guiding the work.
- Compression. How small can a message, a record or a photo become if a model on the device understands what it contains?
- Priority. Which information matters most in a given moment, and can an on-device model decide what to share first?
- Verification. With no server to check against, how can a device confirm that what it received is intact and really came from its sender?
- Privacy and security. How do we build protection in from the start, so that people share only what they choose, and only with the people they choose?
- Affordable phones. Does any of this work on the low-cost phones people actually own, with limited memory and tight battery budgets? Roughly 600 million people in Africa live without electricity, so energy is part of the question.
- Measurement. What counts as success, and how do we test it in the conditions people live in, as well as in the lab?
None of these has a settled answer yet, and some may prove harder than they look. Finding out is the work.
Edge first, by necessity
Without a connection, there is no cloud to call. Anything that compresses, prioritizes, checks or protects what is shared has to run on the device itself, within its memory and its battery.
That is a constraint we already design around. In Small models, big continent, we argued that AI for Africa should work offline by default, run on small models that fit the phones people own and build privacy into the architecture. This research takes that thinking to its limit, where the device is all there is.
Privacy looks different here, too. If data moves between devices with no server in the middle, there is no central store to breach. There is also no server to ask whether a message is genuine, which is why verification is one of our core questions.
Our partners
Questions this hard are better worked on together. Our partners in this research include OpenAI4Africa and Kalon Edge.
OpenAI4Africa, founded in 2023, works under the banner “Borderless AI: for Africans, for all.” It promotes AI research, innovation and development across Africa, with focus areas that run from AI4Health and AI4Education to AI4ClimateAction, AI4Agriculture, AI4Cybersecurity and AI4Finance. It brings together AI experts and enthusiasts from Academia, Government and Industry, the AGI approach that SISU AI’s companies share.
Kalon Edge focuses on edge intelligence: AI that runs on devices, close to where data is created, instead of being confined to the cloud. Its site describes on-device AI that responds faster, is ready to work offline and keeps data private, across mobile, desktop, web and embedded platforms. It says edge AI brings intelligence to devices “so experiences keep working when networks don’t,” a line this research takes literally.
Where it could lead
If this research finds a practical path, the uses could reach well beyond emergencies. A community health worker could bring a colleague’s records up to date after a day spent far from any signal. A teacher could share a lesson with every phone in a classroom that has no connection.
These are possibilities, not promises. Each depends on answers we don’t have yet.
The instinct behind them already shapes our products. Dinera, which applies AI to food safety and food security from farm to fork, has an app built to work online and offline, plus access via USSD. Vaxus, which brings AI to carbon accounting and climate awareness, supports USSD and SMS so people can reach it online or offline. Both channels still rely on a mobile network; this research asks what devices can do with no network at all. What we learn here could one day inform products like these.
What comes next
We will publish a paper on our findings and share it across our platforms. Until then, we would rather ask clear questions in public than make claims we can’t yet support.
Our tagline is “Intelligence without borders.” This research asks whether intelligence can cross one more border: the place where the network ends.