Multi-stakeholder consortium

From Open Standard to Shared Governance

F-ACT is forming a multi-stakeholder consortium — modelled on W3C, IETF, and the Linux Foundation — so the standard that governs autonomous agents is itself governed by humanity, not by any single company.

The standard is already open and royalty-free. This is the next step: putting the pen in the hands of the institutions that must live with the rules. Sign now to join the founding cohort — it closes at 42.

The Call for Signatories is the document to circulate internally — the gap, the standard, the governance handover commitment, and the objections answered, in four pages.

The Founding Circle
/ 42

We welcome signatories continuously, but governance is handed over on legitimacy — not headcount. A big number proves enthusiasm; it does not prove the right institutions are at the table.

The standards race is already on — and nobody has claimed this square

As of 2026 there is still no single, open standard for governing what an autonomous agent is allowed to do before it acts. The gap is being filled piecemeal — by identity protocols, security frameworks, and national initiatives that each solve one corner. F-ACT is the conformance layer that makes any of them checkable.

The regulatory clock is running

EU AI Act enforcement for general-purpose AI began on 2 August 2026, with fines up to €15M or 3% of global turnover. Obligations now exist; a common way to prove an agent met them does not.

Management standards arrived — but stop above the agent

ISO/IEC 42001 was adopted as a European standard (EN ISO/IEC 42001:2026) in March 2026, with national adoption across 34 countries due by September 2026. It governs the organisation, not the individual agent action.

Identity is being standardised; authority is not

The Linux Foundation now stewards the Model Context Protocol; NIST launched an AI Agent Standards Initiative; the Cloud Security Alliance published an Agentic Trust Framework. They answer “who is this agent?” — not “was it allowed to do that, and can you prove it?”

The safety community is converging on agents

The Frontier Model Forum published emerging security practices for AI agents in 2026 and is testifying to legislators. The demand for a deterministic, citable standard is real and rising — the supply is not there yet.

How the handover works

Phase 1: Formation

Now
  • Standard published and freely downloadable
  • Patent pledge: irrevocable, royalty-free, worldwide — covers any use of the core claims
  • Founding Signatories welcomed continuously — the founding cohort closes at 42
  • Editorial pen remains with the originator while the standard stabilises

Phase 2: Shared Governance

Handover
  • Transfers on legitimacy, not headcount — signatory count alone never triggers it
  • Requires 4 of 5 constituencies represented, including regulators and frontier labs
  • And a stable v1.x standard at least 12 months old — no handover of an immature spec
  • Hard backstop: control transfers by 19 Aug 2028 regardless — no indefinite gatekeeping
  • Elected governing board, no single-member veto; originator keeps one seat, no veto

The standard that governs how autonomous agents act on behalf of humanity should itself be governed by humanity.

Become a Founding Signatory

Endorse the principle that agent governance should be open, deterministic, and collectively governed. No fee. No contract. Implementation and formal membership come later.

No fee · No contract · Endorse the principle

This is how durable standards are actually run

A standard that governs everyone cannot be owned by one company. Every standard the internet runs on solved this the same way: the originator hands the pen to a neutral, multi-stakeholder body once the work is stable. We are following the proven playbook, not improvising one.

W3C

The web’s standards are set by working groups under a royalty-free patent policy, reached by consensus across members — no single vendor decides.

IETF

The protocols the internet runs on are published as open RFCs by “rough consensus and running code,” with open participation and no membership fee to contribute.

OASIS

Open standards like SAML and MQTT are governed by a member consortium with an open-participation process — proof the model scales beyond the web.

Linux Foundation

Provides neutral homes for shared infrastructure — now including agent tooling like the Model Context Protocol — so no one company controls what everyone depends on.

Straight answers

What does signing actually commit me to?

Nothing binding. You endorse the principle that agent governance should be open, deterministic, and collectively governed. No fee, no contract, no obligation to adopt anything. Formal membership and implementation are separate, later steps.

Is this just a walled garden with an open label?

No. The standard is already published, free to cite, and covered by an irrevocable, royalty-free patent pledge over any use of the core claims. Any vendor, any stack, can implement it without our permission — and we are measured against the same tiers as everyone else.

Why does the founding cohort close at 42 if handover isn’t about headcount?

Because momentum and legitimacy are different things. Closing the founding cohort at 42 marks a committed founding group. Governance only transfers when the right institutions are actually at the table — a quorum of constituencies including regulators and frontier labs — not when a counter hits a number.

Who controls the standard today, and when does that change?

The originator holds the editorial pen while v1.x stabilises. Control transfers to an elected governing board once the legitimacy quorum is met and the spec is a stable v1.x at least 12 months old — with a hard backstop of 19 August 2028 so it can never be held indefinitely.

What happens to the originator after handover?

One seat on the governing board, no veto — the same as any other member. No single party, including us, can block the standard’s evolution.

Founding Signatories

Loading the roll…