Traefik Labs launches verifiable evidence system for AI agent actions

Overview

Updated Oct 6

AI agents now issue refunds, read customer data, and change business records. When a regulator asks 18 months later who authorized an action and which policy governed it, a standard log entry is not evidence: anyone with access could have edited it.

On September 15, 2026, Traefik Labs introduced the Sovereign Trust Plane (STP) for its Traefik Hub gateway. It connects three functions: delegated access through the customer's identity provider, policy enforcement through the organization's policy engine, and an append-only transparency log that records every approval and refusal. The company planned general availability for September 30, 2026, but has not publicly confirmed that GA shipped on time; as of early October, its product page still lists the components as early access because two underlying standards are moving drafts.

An independent witness cosigns each checkpoint, so verifying the record means checking proofs and signatures, not taking Traefik's word. The stack runs on customer-controlled infrastructure and can operate air-gapped with no SaaS control plane on the request path. Traefik has named no customers or deployments yet.

Why it matters

When an AI agent moves money or changes data, companies must prove who authorized it and which controls fired. That evidence must survive independent review.

Questions about this story

Free account needed to ask — your question is kept and asked for you right after sign-up. Answers are public.

No questions yet — be the first to ask.

Key Indicators

Governance capabilities in STP

Delegate (identity), Authorize (policy), Prove (transparency log).

September 30, 2026

General availability target

Target date has passed. Traefik has not publicly confirmed GA; product page still lists STP as early access as of early October.

December 2027

EU AI Act Article 12 effective date

The EU's record-keeping requirement was delayed to this month, shaping the regulatory deadline.

Voices

Curated perspectives — historical figures and your fellow readers.

People Involved

Emile Vauge

Founder and chief executive of Traefik Labs

Leading the company through the STP launch

Organizations Involved

Traefik Labs

Software company

Launched Sovereign Trust Plane on September 15, 2026

Creates Traefik, one of the most widely used open-source reverse proxies, and the Traefik Hub gateway for API, AI, and model traffic.

Timeline

2015 October 2026

3 events Latest: October 2nd, 2026 · 1 month ago

October 2026
Third-party analysis flags missing customers and independent verification
Latest Analysis

The Clarity notes STP's GA target of September 30 has passed without a Traefik announcement; no named customers, deployments, or external verification of the transparency log scheme have surfaced.
October 2nd, 2026
September 2026
Sovereign Trust Plane introduced
Product launch

STP launched in Traefik Hub with live evidence capability; GA planned September 30. Verified with Okta, OpenFGA, and Cerbos.
September 15th, 2026
2015
Traefik Labs founded
Founding

Company founded as Containous; Traefik reverse proxy released.
2015

Scenarios

STP reaches general availability, regulated industries lead adoption

Likely Resolves by End of 2026

Discussed by: Help Net Security's coverage emphasizes CISO and compliance use cases across banking, insurance, and healthcare.

General availability lands September 30 as planned. Adoption leads in sectors that already face audit requirements, because STP converts questions like 'who authorized this?' into a verifiable record. The pre-verified Okta, OpenFGA, and Cerbos integrations keep the stack on customer-controlled infrastructure, which suits regulated environments. A public Traefik-operated witness covers smaller customers; larger ones self-host.

GA slips, adoption stays confined to early adopters

Possible Resolves by Q1 2027

Discussed by: The EU AI Act Article 12 delay to December 2027 removes near-term regulatory pressure, which may let companies defer the investment.

General availability pushes past September 30. Without a binding regulatory deadline until late 2027, most companies treat tamper-evident agent records as optional. Adoption stays with security-conscious early adopters and the few sectors already under audit obligation. The market waits for regulators to force the issue.

Rival gateways replicate the transparency model before EU rules bind

Uncertain Resolves by End of 2027

Discussed by: EU AI Act Article 12's December 2027 record-keeping requirement creates a market for verifiable evidence.

Competing API and AI gateway vendors ship their own tamper-evident evidence features before the December 2027 EU deadline. The approach becomes table stakes, and Traefik's first-mover edge narrows to its pre-verified identity and policy integrations. Regulators get multiple vendors offering similar evidence, slowing any single-vendor standard.

Historical Context

3 moments from history that rhyme with this story — and how they unfolded.

March 2013

Certificate Transparency (2013)

Google began requiring Certificate Transparency for new HTTPS certificates. Every certificate had to be logged in public append-only ledgers where anyone could audit entries and detect fraudulent issuance.

Then

Fraudulent and misissued certificates became detectable and publicly auditable; browsers began refusing unlogged certificates.

Now

The model, independent verifiers checking append-only logs without trusting the issuer, became a template for building auditable infrastructure.

Why this matters now

STP applies the same pattern to AI agent actions: an append-only transparency log plus an independent witness, so customers verify without trusting Traefik's signature alone.

July 2002

Sarbanes-Oxley Act (2002)

After Enron and WorldCom collapsed on misstated financials, Congress passed Sarbanes-Oxley. Section 404 forced public companies to document and attest that internal controls over financial reporting actually worked, not just report the numbers.

Then

Compliance costs rose sharply as companies built internal control evidence trails.

Now

It created a lasting expectation that executives prove controls operated, and became the template for 'prove the controls fired' regulation.

Why this matters now

STP applies the same idea to AI agents: regulators want to know not only what happened, but what the agent was prevented from doing and which policy produced each decision.

June 2021

SLSA software provenance framework (2021)

Google and the Open Source Security Foundation introduced SLSA, a framework requiring cryptographically signed provenance for software builds so a binary could be traced to the exact source and build process that produced it.

Then

Major cloud platforms adopted signed provenance and attestation for build pipelines.

Now

Verifying software's origin moved from an audit niche to a default expectation in supply chain security.

Why this matters now

STP extends the same 'prove the provenance' standard from software artifacts to agent decisions, recording where authority came from and which policy decided each action.