Who we are
Openwist is an AI research and products company working to build open, safe and accountable AI systems. We do research, build open models and datasets, and make AI products, including Exa, our AI assistant.
Safety and accountability are published alongside the work, in our Safety Framework, our Constitution for Vehicles and Robots, our Accountability Terms and our Usage Policy.
Openwist Limited was founded in 2026 and is based in Nairobi, Kenya. Website: openwist.com. Safety: openwist.com/safety.
Our mission
To responsibly research, build and publish open, safe and accountable AI systems for the long-term benefit of people everywhere.
Five pillars
1. Safe by design
We build safety into our systems from the start, and we test it before release. We treat safety as part of the product, not a step added at the end.
2. Accountable by default
Every system we release has a named person responsible for it. We keep records that let others see what was built, how it was tested and who approved it.
3. Open where safe
We publish models, datasets and research when the benefits outweigh the risks. When we hold something back, we say so and explain why.
4. Safety as shared practice
We publish our safety methods so others can use, check and improve them. We want safe practice to be the norm across the field, not a feature of one company.
5. Benefit for everyone
We build for people everywhere. We aim for AI that is useful, affordable and available well beyond the places where it is usually built.
Our values
People first
Our systems serve the people who use them and the people they affect.
Proportionate caution
The greater the possible harm, the more care and review we apply.
Evidence before claims
We say only what we can show, and we report limits as plainly as results.
Design for understanding
We build systems that people can inspect, question and explain.
Defence in depth
We use several independent safeguards, so that no single failure causes harm.
Traceability
We can trace data, decisions and changes back to their source.
Accountability
We own our results, including our mistakes, and we fix them.
Learn in the open
We share what we learn, including what went wrong.
How this Charter binds us
Our Articles of Association require every director to weigh this Charter when making decisions. Pillars 1 and 2 are intended to join the provisions of the Articles that cannot be weakened without the Founder's approval and the Safety Lead's confirmation. We will report each year on how we have kept to these pillars, together with our annual report on the Constitution for Vehicles and Robots.
Version history
- Adopted by the Board.