Scaling AI Identity: Adding Dutch BSN Support to Soulprint's ZK Standard
These articles are AI-generated summaries. Please check the original sources for full details.
Nederland devs: Add BSN (Burgerservicenummer) to the AI identity standard — Soulprint open source (30 min PR)
Soulprint is an open-source identity standard designed to verify the human origin behind AI agent actions using Zero-Knowledge Proofs. The system utilizes Circom 2.1.8 and Groth16 to achieve 564ms prove times and 25ms verification on-device.
Why This Matters
Current AI agent architectures often lack human-in-the-loop verification, allowing agents to execute financial transactions and sign documents without identity proof. While centralized gatekeepers exist, Soulprint offers a decentralized alternative that runs 100% on-device using MIT-licensed code, preventing Big Tech from becoming the sole arbiter of digital identity. This technical reality addresses the scaling risk of autonomous agents by providing a cryptographic link to human operators without compromising privacy through ZK proofs.
Key Insights
- Soulprint uses Circom 2.1.8 and Groth16 to generate proofs with only 844 constraints (2026).
- On-device performance reaches 564ms for proving and 25ms for verification (Soulprint, 2026).
- The Dutch BSN validation uses a 11-proef algorithm, where a weighted sum of 9 digits must be divisible by 11.
- Integration of new national IDs like the BSN requires a 30-minute contribution to the open-source registry (Soulprint, 2026).
- AI agents can include identity scores (0-100) in HTTP headers using the X-Soulprint tag for server-side tool gatekeeping.
Working Examples
API verification of AI identity score using soulprint-mcp.
import { requireSoulprint } from "soulprint-mcp";
server.tool("premium", requireSoulprint({ minScore: 80 }), handler);
Implementation of the Dutch BSN verifier for the Soulprint registry.
// packages/verify-local/src/document/countries/NL.ts
import { CountryVerifier, DocumentResult, NumberValidation } from "../verifier.interface";
const NL: CountryVerifier = {
countryCode: "NL",
countryName: "Nederland",
documentTypes: ["bsn", "paspoort"],
parse(ocrText: string): DocumentResult {
const doc_number = ocrText.match(/(\d{9})/)?.[1] ?? "";
return { valid: !!doc_number, doc_number, country: "NL" };
},
validate(docNumber: string): NumberValidation {
return { valid: validateBSN(docNumber) };
},
};
export default NL;
Practical Applications
- Use case: AI agents protecting premium API tools by requiring a minimum Soulprint score (e.g., minScore: 80) via requireSoulprint middleware.
- Pitfall: Failing to implement the mod-11 check for BSN, which results in invalidating legitimate Dutch identity documents during OCR parsing.
References:
Continue reading
Next article
RAG vs. Context Stuffing: Benchmarking Efficiency and Reliability in Large Context Windows
Related Content
Agentic OS: A 7-Layer Open-Source Architecture for Multi-Agent Coordination
Mihir N Modi releases Agentic OS, an MIT-licensed 7-layer framework that coordinates specialized AI agents with built-in memory and zero-cost tier support.
Prior Labs Launches TabPFN-2.5: Scaling Tabular Foundation Models for Enhanced Performance and Efficiency
Prior Labs introduces TabPFN-2.5, a major update to its tabular foundation model, enabling handling of 50,000 samples and 2,000 features with no training required, while outperforming traditional models on benchmarks.
LangWatch Open Sources Evaluation Layer for AI Agents to Solve Non-Determinism
LangWatch launches an open-source platform for AI agent evaluation and tracing, enabling developers to move beyond anecdotal testing with end-to-end simulations and OTel-native monitoring.