Submitted to the EU AI Office on 3 June 2026

EU AI Office · Article 50 Consultation

Audact’s submission to the EU AI Office Article 50 consultation (3 June 2026).

Verifiable, machine-readable, jurisdiction-aware compliance proofs as default — submitted to the EU AI Office Article 50 consultation as part of the public stakeholder process.

Five concrete positions in the submission

1. 1 · Verifiability (bewijsbaarheid)

   Article 50 obligations should be operationally verifiable, not declarative. Audact proposes that every AI interaction with a natural person yields a tamper-evident, machine-checkable evidence record covering disclosure, consent, jurisdiction, and timestamp.

2. 2 · Machine-readable proofs

   Evidence chains should be expressible as W3C Verifiable Credentials + C2PA content credentials. This enables cross-vendor auditability and lets Member State DPAs ingest proofs through standard tooling rather than vendor-specific dashboards.

3. 3 · Jurisdiction-awareness

   A per-Member-State policy engine is necessary because Article 50 lives alongside national rules (NL Telecomwet, DE Bundesland-level requirements, FR CNIL guidance). Audact ships a per-MS + per-Bundesland engine and proposes a public, machine-readable rule schema so providers do not each maintain incompatible interpretations.

4. 4 · Provider / deployer contract allocation

   Liability between provider and deployer should be allocated through clear, written contractual terms with default rules in case of silence. Without this, agencies acting as deployers carry uncertain residual exposure for upstream platform behaviour.

5. 5 · Open evidence schema

   Audact publishes its rule schema under AudactRules (CC-BY-SA), an open licence that allows any provider — including competitors — to adopt the same evidence format. Compliance evidence should be portable, not a lock-in mechanism.

Why this matters

Article 50 enforcement begins 2 August 2026. Without operational verifiability, the obligation collapses into self-declared compliance — which is what Member State regulators have repeatedly told the EU AI Office they cannot supervise at scale. The five positions above are designed to be implementable by any provider, not just Audact.