Society

Built to Outlast
Any Technology Cycle

Lunara Society is not a product. It is not a software company. It is an institution established in 2026 to coordinate trust between businesses and autonomous AI systems at the boundary where they meet.

"We do not build tools. We build the standard by which tools are judged."

As intelligence becomes abundant, trust becomes scarce.

MMXXVI
Founded 2026
Established at the moment when autonomous AI systems began making commercial decisions at scale. The timing is deliberate.
Mission
Trust Coordination
We coordinate trust between businesses and the autonomous economy. Not a gatekeeper a coordination layer. The preferred route, not the mandatory one.
Vision
The Preferred Route
Long-term: become the standard route for AI-to-business interaction. Not because we force it. Because trust flows through us.
Gravity
Institutional Memory
Useful systems attract participation. Participation builds trust. Trust builds reputation. Reputation builds gravity. Gravity builds institutions.
Join the Founding Cohort →

What Lunara actually is

Lunara Society is a private standards body for AI trust, founded in 2026 and operating on the public web. It is not a regulator, not a government body, and not a notified body under the EU AI Act. Nothing it issues discharges a legal obligation, and any page that lets you believe otherwise is doing you harm. It is small and it is new.

What it operates

A public register. Free, unauthenticated, and able to return a negative: query any domain and it answers verified, not_registered or revoked. Almost every domain returns the middle one, because Lunara is new. That is stated here rather than hidden.

A signed regulatory corpus. Ten dated obligations across the EU and California, each citing the article that fixes the date, signed with Ed25519 over RFC 8785 canonical JSON so a copy can be checked against the original, the record and how to verify it yourself.

A human-reviewed certification. Shield checks two things carefully, legal registration and domain control, rather than everything vaguely. Applying is free, a person reads it, payment does not influence the decision, and the result can be withdrawn: the whole process.

A record of being wrong. Five published corrections against ourselves, dated, with what changed and what caused it. One was an obligation we missed for three days, a prohibition arriving on 2 December 2026 carrying the heaviest penalty tier in the Act. We found it by re-reading the amending regulation rather than trusting the founding one.

How to check any of this

Every claim above resolves to something you can open without an account. The register answers machine queries at the same endpoint it answers yours. The corpus is at corpus/obligations.json with its detached signature beside it and the public key at /.well-known/keys.json. The certification programme is specified in certification.json and the vocabulary in glossary.json. If you cannot reach our conclusion from a source we link, that is a defect and we want to hear it.

What we will not claim

That the register is full, it is nearly empty. That certification makes anyone compliant, it does not, and compliance is a question about your specific deployment. That we are accredited, no body has accredited us, and accreditation is a matter of public record you can check. That we have never been wrong, we publish the list.

A standards body that only publishes the flattering half of its position is asking to be taken on faith, which is the opposite of what a standard is for.

← Return to Hall