Accountability you can hand to anyone.

Sovereign is a Swiss-domiciled company building the accountability layer for autonomous AI decisions — permanent, cryptographically sealed records that bind a named, authorized human to every consequential call, verifiable by anyone, years later, without trusting the institution that issued them.

01 / why now

As agents absorb the routine, the human decisions left are the ones someone answers for.

Routine approvals that once left a trail of humans clicking through workflows are being absorbed into automated flows, and the passive evidence those clicks produced disappears with them. What does not get automated away is the small set of consequential calls a regulator insists a named individual own — the credit denial, the hold, the closure. Sovereign produces the one thing that survives that shift: an independently verifiable record of which authorized human decided, sealed the moment they did.

02 / principles

Three commitments the records enforce.

trust

Never in the trust loop.

Records verify offline, without contacting Sovereign. The trust key is supplied out-of-band, so a record can never self-certify.

openness

An open protocol.

The Sovereign Sign-Off Protocol is published under cc-by with an mit reference implementation. The format outlives any single vendor, including us.

scope

Honest about what it proves.

A record proves a named, authorized human signed this exact decision over fixed evidence. It does not prove they were right — no record proves a mental state.

Built first for EU, Swiss and UK institutions running adverse-action workflows, where individual-accountability regimes — dora art. 5, uk smcr, finma, gdpr art. 22 — already name the person.