Accountability record

This company runs on AI. Here is the record.

Trusq is operated by a small team with automated systems doing most of the work: monitoring the primary EU sources, drafting, publishing and re-checking. That is the direction a lot of organisations are heading. It creates one new problem — when software does the work, who is still accountable for what it says? This page is our own answer, in numbers, including the parts that are unfinished or wrong.

2 965automated source checks
61changes detected
26published
1awaiting human verification

Live counts from the same database that serves the updates feed. Publishing since 26 Jun 2026. Most recent entry 3 Aug 2026. Of everything detected, 53 arrived through automated monitoring rather than a person adding it, and 25 published entries have since been checked by a human against their primary source.

Publish first, verify in the open

A change detected in an official EU source is published immediately, labelled not yet editorially verified, with a link to its primary source. A person reviews it afterwards and refines or retracts it. We label the state rather than hide it: the count of unverified entries above is deliberately on this page, because a number that can only go down looks like marketing, not like a control.

A control that is allowed to fail

Most monitoring proves a system is running. It does not prove the system is still right. A separate hourly control re-checks our own published claims against the official sources; when a claim no longer holds it turns RED, the service unit goes to failed, and that is the alarm — there is no dashboard to quietly ignore.

Most recent run: RED — 13 green, 1 red, 0 amber, 0 unknown. Last run 2026-08-10 03:00 UTC.

What we got wrong, and fixed in public

Automated publication means mistakes reach the page before a person has read them. That is a deliberate trade: speed on a deadline is worth more than a queue, provided every correction is public and the original URL never moves. 3 published entries have been corrected after publication.

EntryWhat was wrong
Commission fines AliExpress €550 million for breaching the Digital Services ActFiled under the wrong framework and described as enforcement against a member state. AliExpress is a platform, and the action falls under the DSA, not the AI Act.
Commission publishes Code of Practice on marking and labelling AI-generated…Mapped to the GPAI articles (53/55). It actually supports the Article 50(2)/(4) transparency duties.
Digital Omnibus — high-risk deferralStatus was stale: still recorded as pending after the act had been published in the Official Journal on 24 July 2026 and entered into force.

What a human still decides

Automation detects, drafts, publishes and re-checks. It does not decide what counts as authoritative: the active rule corpus is released by a person through an explicit, logged release step, and every release is recorded in an audit trail with who did it and when. Customer content is processed only by an EU AI provider; no customer data reaches a US model. The boundaries are on our trust & data page.

Why this matters beyond us

If your organisation is heading the same way — more output, fewer hands, software in the middle — you will be asked the same questions by a customer, an auditor, an insurer or a regulator. Under the AI Act's transparency duties, in force since 2 August 2026, some of those questions are now legal ones. The record on this page is the artefact we think every AI-operated company will need. You can see the same machinery applied to EU regulation on how Trusq stays current, or check one of your own systems with the free Article 50 check.
← Back to Trusq