Prepared for Caroline Parker — private link, please don't circulate
VAC Protocol
Standards & intellectual-property — a note for your read · Violet Shores Pty Ltd
Two competent people, verified live — at the moment of the decision.
VAC proves each is competent (authority), genuinely present (live biometric), and that both were there (immutable audit). A two-person control built on logins can prove none of that.
The standards move, in plain terms
The short version. The idea at the heart of VAC — proving which real, accountable person stands behind an AI system's actions — lines up with an active effort by the global body that writes the internet's technical rulebooks to create shared standards for what AI agents do. That effort has an obvious gap our idea fills. We are moving on it now: two draft documents describing the outward interface — deliberately not the inner workings — are finalised and being submitted this week, timed to the body's meeting in Vienna (which I'm attending). The interface goes public to shape the field; the mechanisms stay protected by the patent filings.
A few plain-language definitions
- IETF (Internet Engineering Task Force)
- The open, global body that writes the internet's core technical standards — the agreed rules that let different companies' systems work together. Membership is free and open to anyone.
- Standard
- An agreed technical specification. Following a recognised standard is how a company shows its product "plays by the rules" and works with everyone else's.
- Draft (Internet-Draft)
- An early proposal document submitted to the IETF. It has no official status on its own and expires after six months — it's a starting point for discussion, not a finished standard.
- Working Group
- The committee inside the IETF that, if it adopts a draft, shepherds it over several years toward becoming a finished, published standard.
- IPR disclosure
- An IETF rule requiring anyone contributing to declare relevant patents or patent applications they hold, along with the terms on which they'd license them.
Where things stand this week
Two provisional patent applications are filed (references AU 2026906147 and AU 2026906148), covering how the technology works. Two Internet-Drafts covering only the outward interface are being submitted to the IETF this week, and I'll be at the IETF's Vienna meeting where the agent-protocols conversation is happening. The posture is compose, don't contest: a constructive presence that positions our work as the missing accountability layer alongside what others are building — while the mechanism patents stay private and separately licensable.
Decided · for your awareness
Why publish now rather than wait for a signed customer
The calculus flipped: the agent-standards conversation is happening this week, with or without us. The drafts disclose only the interface (which we want adopted), never the mechanisms (which the filings protect) — so absence now costs field-shaping influence and credibility, while adding little protection. If anything in this reasoning worries you legally, I genuinely want to hear it on our call.
For your awareness — no homework, just things I’d love your reaction to when we talk
On my mind · 1 — the licensing statement
The IETF asks contributors to declare patents and say how they'd license them.
My intent: keep the interface free-to-use and open (drives adoption, suits the Foundation), while the underlying mechanism patents stay separately licensable. That split feels right to me — if your instinct says otherwise, that's exactly the reaction I want over a coffee-time call, nothing more.
On my mind · 2 — the conversion timing
The provisional-to-full patent conversions fall due Feb/Mar 2027, with the interface now public.
I'm thinking through which claims to convert and broaden. Purely FYI at this stage — I'll bring it properly when it's actually a task, months from now.
Relevant to your world
This isn't only an EU idea. Australia's own AI guidance (AI Adoption / "AI6", Practice 3) requires meaningful human oversight for high-stakes decisions — "rubber-stamping is not oversight" — and since June 2026 government bodies must name an accountable human for every AI use (tribunals are public bodies). The EU AI Act legislates it directly — Article 14 (human oversight, two-person verification) and Article 12 (record-keeping) — for high-risk systems from 2 December 2027. New Zealand and the Five Eyes agencies point the same way. VAC is built to evidence exactly this.
AU · AI6 human oversightAU gov · accountable owner (binding)EU Art 14 · two-personEU Art 12 · records