Add-on module

AI Governance

Every AI system you use needs an owner. And someone who can stop it.

AI governance software for the EU AI Act. AI Governance in Acuna GRC keeps one register of every AI system you build, buy or deploy, gates its approval on an impact assessment, names who can stop it and records what goes wrong. Aiko lists the open gaps, and Swiss experts can assess your AI use with you.

  • 35+ years of experience
  • 50+ frameworks
  • ISO 27001 certified

The problem

Why does AI use slip past governance? Nobody keeps the list.

Teams buy AI tools, switch on AI features inside software you already use, or build their own, and nobody records who checked them. When an AI decision goes wrong, nobody knows who can stop it.

  • Nobody can tell me which AI tools we actually use.

    Fix: one register of every system you build, buy or deploy, with its provider and what it does.

  • A team switched on an AI feature with no review.

    Fix: Approved for use needs an approved impact assessment or a screening decision first, and "in production without approval" shows as a warning.

  • If an AI decision goes wrong, who can stop it?

    Fix: overseers are named with the authority to override or suspend, and the Nobody can stop it lens lists the gaps.

How AI Governance works

How does AI governance software work in Acuna GRC? Register, assess, approve, oversee.

You register each AI system, screen it and complete an impact assessment where one is owed, approve it for use, then name its overseers and record incidents and contestations against it.

01 · AI Systems

How do you build an inventory of the AI systems you build, buy or deploy?

With New AI System. Each record gets a reference such as AIS-001 and holds what the system does, its provider, lifecycle stage, autonomy and who it affects. The Classification tab records your role, the risk class, the use case and a required rationale. Five warnings, such as "no one can stop this system", show what is missing and clear when you fix the fact.

  • Risk tier and Assessed risk
  • Five derived warnings
  • Linked assets, risks and controls
AI providers are suppliers you assess in TPRM →
Try it · EU AI Act, Article 6

How do you classify an AI system's risk under the EU AI Act?

Start from its intended purpose. Answer up to four questions to see a likely risk class and the rationale the Classification tab would keep.

EU AI ACT · ARTICLE 6

Is your AI system high-risk?

Four questions at most, in the order a regulator asks them. Nothing is saved and nothing leaves your browser. This runs entirely on this page.

An orientation tool, not legal advice. Classification turns on your system’s actual intended purpose and the documentation behind it. Confirm the result against the Act and the Commission guidelines.

02 · Impact Assessments

When is an AI impact assessment required, and what goes in it?

Screening decides. The Screening tab asks whether the system makes or materially informs a decision about a person, and whether children or vulnerable groups are affected. If an assessment is owed, a five-part wizard covers Purpose & context, Affected parties & harms, Oversight & contestation, Measures & residual risk, and Review. Tick Fundamental-rights impact assessment and one more section appears. Approval freezes it, and Generate report files a Word document.

  • Dated "Not required" decisions
  • FRIA in the same record
  • Review date and task
Processing activities and DPIAs in Data Privacy →
03 · Human Oversight

Who can stop an AI system, and how is that recorded?

The overseers you name. Each assignment records a person from your organization, the kind of oversight, whether they can override an outcome or suspend the system, a competence basis and an effective-from date. Both authority boxes are off unless ticked. The No overseer at all and Nobody can stop it lenses gather the gaps.

  • Can override
  • Can suspend
  • Ended, never deleted
Who owns AI governance, and the Friday test →
04 · Incidents and contestations

What happens when an AI system causes an incident, or someone contests a decision?

It runs on a clock. An AI incident records the severity, who was affected, the root cause and an owner. A Serious one gets a notification due 72 hours after discovery, and every incident opens a linked issue. People affected by a decision use your public Contest an AI-assisted decision page, with no account. Each contestation gets a reference such as AIC-2026-001 and is due 30 days after receipt.

  • 72-hour clock on Serious incidents
  • Linked issue every time
  • Raise incident from a contestation
Webinar: From AI hype to AI reality →

Aiko in AI Governance

What does the AI assistant do in AI Governance? Aiko lists what is missing.

Ask Aiko "What is missing for my AI conformity?" and it reads your AI system register and lists the open gaps, within your access profile. Your people decide what to fix.

You set the paceSuggest onlyAsk me firstdefaultDo it and tell meSome things always ask first, and every step goes into the Activity log. How Aiko works

Swiss experts

Can someone run an AI gap assessment for us? Yes, with our Swiss experts.

Done with you: our experts assess your AI use against the EU AI Act and ISO 42001 together with your team. Your people keep the register and make the decisions.

  • Jean-Emmanuel Rodriguez

    Cybersecurity governance expert

    Jean-Emmanuel Rodriguez

    AI governance
  • Laura Menetrey

    Legal and compliance expert

    Laura Menetrey

    Legal and privacy
  • Alexis Hirschhorn

    CEO of Acuna

    Alexis Hirschhorn

    Certified Lead Auditor
  1. Do it yourself

    You run the program in Acuna GRC, with Aiko.

  2. Done with youOn AI Governance

    Guidance, reviews and audit preparation with your team.

  3. Done for you

    Our experts run it. You review.

Integrations

Which tools does AI governance software connect to? The ones that hold your people, tasks and suppliers.

Review tasks sync both ways with Jira and ServiceNow, the people you name come from your Microsoft Entra directory, and scheduled imports and the API keep the assets and providers behind each AI system current.

  • Tickets and tasks

    Jira and ServiceNow

    Review tasks and other AI governance tasks stay in step both ways: title, status, assignee and due date. Close the ticket and the task closes.

  • Directory

    Microsoft Entra

    Admins invite the people you name as owners and overseers by searching your Microsoft directory, without syncing users.

  • Scheduled imports

    CMDB or procurement

    Your system pushes a CSV of assets and third parties, so the assets an AI system runs on and its provider stay current, matched on an External ID.

  • API v1

    Your own systems

    Create and update the tasks, assets and third parties linked to your AI systems with a scoped key, and get a signed webhook when they change.

  • Jira
  • ServiceNow
  • Entra Directory
  • Entra ID
  • Okta
  • Google Workspace
  • Microsoft 365
  • Azure
  • AWS
  • Google Cloud
  • 100+ connectors in all

Missing a connector? We build it. See Integrations and API

FAQ

What else do people ask about AI governance software?

Does Acuna GRC certify us under the EU AI Act or ISO 42001?

No. AI Governance keeps the operational records those obligations ask for: the register, impact assessments, oversight, incidents and contestations. It does not certify conformity with any law or standard.

Is AI Governance the same as Aiko?

No. AI Governance is the register of your own AI systems. Aiko is the assistant built into Acuna GRC that can read that register and answer what is missing.

How do we check what our AI provider trained on?

With Data Provenance. The questionnaire you send a provider about its training data, quality and safeguards is an ordinary supplier assessment from TPRM, linked to the system. Each system shows Collected, Stale or Missing.

Can a DPIA and an AI impact assessment sit together?

Yes. Link the AI system to its processing activities in Data Privacy, so a DPIA and an impact assessment sit on the same record. The fundamental rights impact assessment, explained.

Can we export an audit trail?

Yes. Export audit trail gives a CSV of every change across impact assessments, oversight assignments, incidents and contestations for the date range you choose.

What does a person see on the contestation page?

A form titled Contest an AI-assisted decision, at your own public address and with your branding. It never lists your AI systems or shows anything from your register. Staff link the actual system afterwards.

How do we get AI Governance, and where is our data hosted?

It is an add-on that Acuna switches on for your organization, and each person's access level decides what they see. Your data is hosted in Switzerland, the EU or the US.

Start with one step.

Bring one AI tool your teams use. We'll register it with you, run the screening and show you what the impact assessment and oversight record.

Hosted in Switzerland, the EU or the US