Home / Add-ons / Compliance Audit
Compliance audit — a prioritised gap report for AI Act, NIS2 or EAA
A scoped technical assessment: what applies to you, what is missing, and the order to fix it. A report and an evidence appendix, not legal advice.
Who it is for
Teams that have been asked for an AI Act, NIS2 or accessibility position and need a defensible starting point.
The problem it removes
- A customer or authority asked a question nobody can answer.
- The obligations are spread across teams and none of them own it.
- Consultants quote five figures before saying what is missing.
What you get
- Scope workshop: systems, roles, jurisdictions
- Gap register mapped to the framework you name
- Prioritised remediation plan with effort estimates
- Evidence appendix you can attach to a response
- One review call after delivery
How to use it
1. Scope
One workshop to list the AI systems and who owns them.
2. Assessment
We map each system to the obligations that apply and record the evidence.
3. Report
A gap register, a remediation order and an evidence appendix.
What the assessment covers
The audit is scoped to one framework, named at intake, and to the AI systems you list in the workshop. Trying to cover three frameworks at once produces a document that is precise about none of them. If you need more than one, they are run as separate engagements with separate registers.
| Framework | What is mapped | What you receive |
|---|---|---|
| EU AI Act | Your role (provider or deployer), Article 50 transparency duties, general-purpose model dependency, prohibited-practice screen | System register entries and a disclosure gap list |
| NIS2 | Article 21 measures that touch your AI systems, the incident reporting path, supplier exposure | A measure register with evidence status and runbook gaps |
| European Accessibility Act | Customer-facing interfaces that consume or present AI output, evidence against WCAG 2.2 AA | An accessibility gap list ordered by user impact |
The workshop is one session of up to 90 minutes. It produces the system list, the owners, and the jurisdictions — the three inputs everything else depends on. Systems nobody can name do not get assessed, which is itself a finding worth recording.
What counts as evidence
A gap register is only useful if each row can be closed with something a reviewer can open. The audit distinguishes a control that exists from a control that can be evidenced, because the second is what an authority or a customer asks for.
- A policy with an owner and a date — not a wiki page last edited two years ago with no owner.
- A ticket or log extract showing the control operated, not just that it was written down.
- A screenshot or recording of the user-facing disclosure in the place a user actually sees it.
- A contract clause covering a supplier's obligation, with the clause identified.
- A training record with names and dates, where AI literacy obligations apply.
An assertion in a meeting is not evidence. The report marks those rows as unevidenced rather than closing them, which is the honest outcome and the one that survives a follow-up question.
How the gap register is scored
Every row gets a severity based on whether the obligation applies and whether evidence exists. Severity drives the remediation order, so the first thirty days of work are the rows that carry the most exposure.
| Severity | Meaning | Target |
|---|---|---|
| Critical | Obligation applies, no control in place, no evidence | 30 days |
| High | Control exists but cannot be evidenced | 60 days |
| Medium | Partial control, or evidence only for part of the scope | 90 days |
| Low | Documentation polish; no exposure if left | Next cycle |
Effort is estimated in days for each row, so the plan can be resourced rather than admired. Where an obligation depends on a supplier, the row names the supplier and the clause to request.
What this is not
It is not legal advice, not a certification, not a conformity assessment, and not a regulator submission. Those are separate activities with separate professional requirements, and no technical assessment can substitute for them. What the report gives you is a defensible description of your current position and a prioritised list of what to fix — the work that has to happen before a lawyer, an auditor or a customer's security team can be answered properly.
We will not describe the output as compliant, certified or approved, and we ask that it is not marketed that way either. A gap register that says "critical, open, 30 days" is more useful than a badge.
Timeline
| Day | Activity | Output |
|---|---|---|
| 0 | Intake: framework, systems, jurisdictions | Confirmed scope |
| 1 | Scope workshop, up to 90 minutes | System list with owners |
| 2–4 | Evidence review and gap analysis | Draft register |
| 5 | Report and evidence appendix | PDF, register, appendix |
| Within 14 days | One review call | Answers on the report |
Which framework to start with
If more than one applies, start with the one a customer or authority has actually asked about. A register written for a question nobody asked is hard to defend and harder to resource. When there is no external prompt, the usual order for a small team is the AI Act first, because the transparency duties are the ones most likely to be visible to end users and the easiest to evidence quickly; NIS2 second, because the incident path has to exist before an incident; accessibility last, because remediation tends to be engineering work rather than documentation.
The order is a recommendation, not a legal position. Your counsel may see it differently, and when they do, their order wins — the assessment is built to be re-run against whichever framework you name next.
What happens after the report
Most teams use the register as a work plan: the critical rows become tickets, the high rows become a quarter's backlog, and the low rows wait. The review call exists to answer questions about the report, not to sell the next engagement. If the honest next step is a lawyer or an accessibility specialist, the call will say so.
Registers go stale. The report is dated, and a register older than a quarter should be treated as a historical document rather than a current position. Re-running the assessment after a material change — a new AI system, a new market, a new supplier — is cheaper than defending an out-of-date one.
Why it costs $500
The engagement is one workshop, three days of review, a report and a call. $500 is deliberately below what a consultancy charges for a scoping conversation, because the point is to let a small team start rather than to maximise the invoice. It is not a substitute for a full conformity assessment, and it is not priced like one.
Questions
Is this legal advice?
No. It is a technical and operational assessment. Your counsel decides legal sufficiency.
What does it cost?
$500 one-time for the scoped assessment and report.
How long?
Five business days after the workshop.
Which frameworks?
EU AI Act (including Article 50), NIS2 and the European Accessibility Act. Name one at intake.
What do we receive?
A PDF report, a gap register in a spreadsheet, and an evidence appendix.
Sources
- Regulation (EU) 2024/1689 — Artificial Intelligence Act — checked 21 November 2026
- Directive (EU) 2022/2555 — NIS2 — checked 21 November 2026
- Directive (EU) 2019/882 — European Accessibility Act — checked 21 November 2026
Related add-ons
Guided Onboarding
Two live sessions with a Plugsky engineer: keys, first integration, an evaluation set and a cost baseline. Delivered in one week.
Migration Service
A compatibility review, a shadow-traffic run and a quality diff before cutover, so you change the base URL with evidence instead of hope.
One intake form, then two sessions in the same week. Paid once through Stripe; nothing renews.
Book the engagementCompare add-ons