The thinking behind the architecture.
Long-form analysis on AI governance, trust architecture, regulatory compliance, and standards positioning, from the people building it. Full Ed25519-signed aiGEC™ certificates with externally verifiable JWKS ship in v2.5.
Working paper.
Our inaugural long-form publication is a working paper for the standards-body and policy audience working on EU AI Act compliance ahead of the EU AI Act high-risk implementation timetable, subject to final adopted AI Omnibus amendments. Open access under CC BY 4.0 on Zenodo.
Closing the EU AI Act Article 12 Logging Gap
A technical specification for per-inference governance certification.
Identifies four architectural properties, per-inference, architecturally bound, externally verifiable, fail-closed, as the minimum required for Article 12-compliant logging. Generalises the Integrity Clash argument (Nemecek et al. 2026) to the audit-versus-enforcement boundary. Presents per-inference governance certification as a candidate technical specification. Targets BSI ART/1, ISO/IEC JTC 1/SC 42, and EU AI Act Article 12 compliance leads.
Future posts on this blog will cover topics including the architectural design of the trust stack, the technical case for per-inference governance certification, regulatory analysis of the EU AI Act and adjacent frameworks, standards-body engagement learnings, and case studies from regulated-industry deployments. Each post will carry a full Ed25519-signed aiGEC™ certificate in v2.5, the same architecture we sell, applied to our own publications.
How we publish.
We treat the blog as part of the public record. Every post is dated, attributed, and signed. Material claims are sourced. Patent positions and standards positions are stated with reference to filings and FRAND undertakings. Industry references are cited; quotes are attributed; framings of other parties' work are accurate and respectful.
Where a blog post relates to a regulated topic (financial services, healthcare, public sector, AI governance enforcement), we are clear about the basis on which we are commenting, as holders of patent applications, standards-positioning contributors, technology providers, or commentators, and where applicable we declare commercial interests.
Where readers identify factual errors, we correct them transparently. Corrections to substantive claims are flagged at the head of the affected post and noted at the foot. Once the v2.5 aiGEC™ certificate pipeline is live, the certificate is re-issued for the corrected text; the original certificate remains discoverable and the chain of issuance is auditable.
For new long-form publications.
Standards-body engagers, policy researchers, regulators, and industry commentators can subscribe to new long-form publications by making contact. Each post is also announced on our news page.
Contact us