Workstream 01
CAP-AI identity, delegation, admission, evidence, and recovery contracts
AI Labs FutureWorks Research profile
Executable assurance requirements for identity, authority, evidence, privacy, and recovery in agent systems.
A Compacitas research initiative
Genesis Governance and AssuranceAI assurance
The Genesis assurance work develops an explicit contract for the complete action path of an AI system: its identity, permitted scope, model admission, execution evidence, review, and recovery. CAP-AI defines the technical profile, while CIV levels specify increasingly demanding assurance conditions. The accompanying privacy and federated-verification work studies how assessors can verify a relevant claim without demanding unrelated raw prompts, private data, or secrets. These documents form one connected initiative, with open questions on independence, common-mode failures, offline evidence, and proportionate review.
The research question
Research workstreams
CAP-AI identity, delegation, admission, evidence, and recovery contracts
CIV assessment levels, independent evaluation, and adversarial conformance
Privacy-preserving evidence and federated verification
Evidence priorities
The initiative is framed around testable questions. Its research direction should be assessed against observed evidence and the limits of each experiment.
Research goals and proposed mechanisms are not established findings. Product-linked tracks describe the research behind a capability, rather than a claim that its hypotheses have been validated.
Associated company
The verifiable civilization runtime for model governance, scheduling, optimization, and authorized compute federation.
Explore Compacitas
MultiplexCompacitas · Prototype
Scheduling many persistent agent identities across bounded model residency and physical compute.
AletheonCompacitas · Protocol in development
A controlled emergence ecology for testing persistent identity, history, diverse substrates, and governance needs.
Verifiable Civilizational LearningCompacitas · Proposed study
Testing whether verified experience can accumulate outside model weights and survive complete agent and model turnover.