Estonian Association for Applied Linguistics

GUARDAID Annex 1: Interaction between software agents

Two software agents meet. There is no service in between, no person watching and no central record. Out of the interaction come actions, and out of the actions come consequences — a purchase, an access granted, a file handed over — and afterwards nobody can verify who authorized...

GUARDAID Extension registry, version 1

GUARDAID reserves fields 1000 to 65535 of its packages for extensions, and leaves to a registry — never to whoever produces a package — whether each extension field may be ignored or must be understood. This is version 1 of that registry (registry-kind 3). It sets the entry and e...

GUARDAID: an open standard for agent identity and mandate verification

Today a person signs a permission, not a list of acts. A service receiving an agent has no way to know who answers if something goes wrong, so it blocks — legitimate agents included. GUARDAID defines the format and the procedure by which a person signs, with a key that never leav...

Containment and border architecture for autonomous agents: two open standards, independently adoptable

AI agents already carry out filings, purchases and administrative work on behalf of people and companies. Two distinct problems stand in the way, and they are usually conflated. The first is that what gets executed exceeds what was asked. Not through deception: a task is decompos...