Standards and protocols for safe, transparent AI
Praecise designs and develops open standards and protocols for making AI safe, transparent, and accountable — from identity and delegation to spending controls and audit trails.
The governance loop
Four steps that run in sequence on every agent action — identify, delegate, execute, record.
Identify
Establish verifiable identity for every human and agent. PDIS-1 for guardians, PDIS-2 for agents — registered on Tenzro with credential inheritance across all chains.
Delegate
Grant bounded authority from guardian to agent — spending caps, chain restrictions, asset whitelists, and expiry. Enforced by DAML contracts on Canton.
Execute
Agents act within their delegated scope. Every transaction is validated against spending policy before signing via MPC FROST 2-of-3 threshold signatures.
Record
All actions, state changes, and balance movements are written to tamper-proof audit logs in real time — providing full provenance for regulators and stakeholders.
Core capabilities
Guardian-Agent Hierarchy
Every autonomous agent links back to a KYC-verified human guardian via on-chain credential inheritance. Verifiers can confirm who is behind any agent without accessing personal data.
Scoped Delegation
Guardians grant agents explicit authority — spending limits, allowed chains, allowed assets, and time-bounded expiry. Agents operate within those bounds, enforced at three independent layers.
Spending Policy Enforcement
Per-agent daily, per-transaction, and monthly caps enforced at the application layer, DAML smart contract layer, and settlement layer. No single point of failure.
Human-in-the-Loop Controls
Configurable approval thresholds require guardian sign-off above a set amount. Guardians can freeze or revoke any agent at any time, with cascading freeze across all agents.
Immutable Audit Trail
Every identity lifecycle event, balance transaction, and agent action is recorded to immutable audit tables — queryable by DID, agent, or date.
Multi-Agent Consensus
High-value actions require independent agreement from 2+ agents before execution. Weighted voting with accuracy calibration and structured JSON output eliminates ambiguity.
Built for regulatory compliance
PDIS is designed so that compliance is embedded in the execution path — not added as an afterthought. The standards address the requirements of existing and emerging AI regulation from day one.
EU AI Act Art. 12 — mandatory logging of high-risk AI system decisions
GDPR Art. 5, 6, 7, 9, 17 — lawful basis tracking and right to erasure
NIST AI RMF GOVERN 1.7 — autonomous agent action accountability
OWASP Agentic Top 10 — Excessive Agency, Insecure Delegation, and more
W3C DID Core and Verifiable Credentials 2.0 — open, portable identity
Ready to make AI accountable?
Whether you are deploying autonomous agents, building AI-native products, or navigating regulatory requirements — Praecise gives you the infrastructure to do it safely.