Making AI systems worthy of reliance
Cortex Foundation builds Cortex Chat, Cortex Code, and Cortex Bot. We are a research-first lab: useful systems, and the science to know how they behave.
Frontier AI will shape every institution people depend on. We believe that transition should be studied, measured, and stewarded — not left to chance.
Understanding first
We do not deploy what we cannot explain. Interpretability research precedes — and gates — every product release.
Evidence over intuition
Claims about model behavior are treated like scientific hypotheses — pre-registered, measured, and published, including when we are wrong.
Safety as a discipline
Safety here is an engineering practice with tests, limits, and failure reports — not a line in a mission statement.
Shared benefit
Cortex is for the people who use it. We publish what we can about how the systems work, rather than keeping the science behind a closed programme.
- 01
Interpretability research precedes — and gates — every product release.
- 02
Every capability we ship is evaluated for misuse.
- 03
Research notes appear on this site when we publish them. Nothing here is a paper we have not written.
- 04
A Cortex Security review that could not run is a failure, never an approval.
- 05
Compute follows the question, not the demo.
- 06
If a question embarrasses us, we still answer it in public.
What we build
Cortex Foundation builds Cortex Chat, Cortex Code, and Cortex Bot. We are a research-first lab. There is no named board, independent trust, or public-benefit corporation on record here to list.
Read the research programme