Verifiable human authority over AI — held in Europe, not in a foreign cloud.
AI agents are starting to take irreversible actions — moving money, changing official records, cutting off services. EMILIA is the open layer that proves a named, accountable human authorized each one, checkable by any European institution, offline, without trusting the operator that produced it.
It turns EU AI Act Article 14 from a principle into a verifiable artifact. Europe holds the proof, not Big Tech.
Europe mandates human oversight of AI. Nothing yet proves it happened.
When an autonomous system acts, the record that a named human authorized that exact action — at the right scope, currently, under the right authority — is today a log the operator keeps and could alter, backfill, or rubber-stamp. That operator is often a non-European hyperscaler. The accountability exists on paper, but not as an artifact anyone independent can check.
Authorize → gate → verify.
Four things that make this undeniable.
EMILIA proves authorization — not that a decision was wise or lawful.
It is a necessary, not sufficient, condition for trustworthy AI — the verifiable foundation other safeguards build on. And EMILIA does not charge public institutions for the core protocol: the goal is adoption of the standard, in the public interest.