someone in rotterdam types a question they wouldn’t say out loud. gets an answer. moves on.
summa is what happens if that sentence does not have to end that way: an eu-hosted, sovereign ai platform that does not train on what you send it, and does not charge extra to be sure of that. it is not open yet — join the waitlist to be first in.
scroll to begin
the useful system sees more than a prompt.
a team runs a supplier contract through an ai tool before it ships. the liability clause gets flagged — sharp, precise, useful. what the team doesn't see is the contract also entering a training pipeline, six time zones away, under a jurisdiction no one agreed to.
the same review, reasoned from what's actually authorised.
summa reads the same clause with the same precision. the difference is structural, not technical — the document leaves no trace unless you choose to keep one.
if you buy something from a company whose real business is selling your attention, you didn't buy a product. you became one.
we charge for access, not attention.
that's not a value statement, it's a business model, and the business model is the actual guarantee.
capable, without surrendering anything.
privacy was never the reason ai got worse. those got tangled together by companies that needed your data to make the business model work — not by any real technical trade-off. summa runs on the same class of models as everyone else.
the only difference is where they run, and what happens to what you told them.
summa is not open yet. it will be, soon, for the people who asked first.
request early access →