AIUC-1 readiness for New Zealand companies that sell AI agents
AIUC-1 is the certification written for an AI agent rather than for the company behind it, and Australian, US and other enterprise buyers now ask New Zealand vendors for it the way they ask for SOC 2. Aegentra takes a company that sells an agent from “we think we comply” to a documented, evidenced, test-ready position against the Q4 2026 release, from NZ$8,125 per agent, then hands over to the AIUC-accredited auditor you appoint. We prepare; independent accredited bodies certify.
Per agent, for one AI agent with a single clearly bounded purpose, one legal entity, existing core security controls and an internal owner able to complete agreed actions. Two agents on a shared platform are quoted together after scoping; a wider or less bounded scope needs a wider quote.
Inclusions and exclusions
Included in the agreed scope
A scoping workshop and a gap assessment dated to the current AIUC-1 release
A requirement-to-control-to-evidence map with owners and priorities, and tailoring of the agreed policies and ownership records
Crosswalk of existing ISO 42001, SOC 2 or ISO 27001 evidence through AIUC’s published mappings
One pre-audit readiness review, one test-readiness briefing and one handover meeting with the accredited auditor you appoint
Separate or excluded
The accredited audit, the adversarial technical testing and the AIUC-1 certificate
Engineering changes to the agent, its tools, connectors or guardrails
Additional agents, legal entities or use cases
Support through quarterly retesting, the 12-month term and further review cycles
Prices in NZD, excluding any applicable tax. Your written proposal confirms the scope, tax treatment and total payable before you approve the work.
Starting prices and indicative ranges are for the stated scope, not an unconditional fixed fee. Other services are priced by scoped quote. Academy course prices are separate from consulting, audit and technical-service fees.
On this page +
Start with the boundary
Define the work that is needed.
For New Zealand companies that build or resell an AI agent and have been asked for AIUC-1 by an enterprise customer, a prospective Australian or US account, or a CPA firm planning a joint SOC 2 and AIUC-1 engagement.
The agent boundary: product, model, tools and connectors, data, guardrails and the buyer question it must answer
Gap assessment against every requirement of the Q4 2026 release, dated to that release
Policies, ownership records and the evidence pack an accredited auditor samples
Crosswalk of existing ISO 42001, SOC 2 or ISO 27001 evidence, test readiness and auditor handover
Keep the responsibilities distinct
Aegentra provides readiness only. Aegentra does not certify against AIUC-1, does not audit for AIUC, does not run the accredited adversarial testing and holds no AIUC accreditation. AIUC-accredited auditors and testing bodies are independent third parties; AIUC’s certification committee decides. Neither a certificate nor a certificate date is guaranteed.
What AIUC-1 is
The standard, as it stands at the Q4 2026 release.
AIUC-1 certifies a single AI agent: the product your customer switches on, together with its model, the tools and connectors it can reach, the data it touches and the guardrails around it. The Artificial Intelligence Underwriting Company (AIUC) published it in July 2025 with contributors from Stanford, MITRE and industry. That makes it a different kind of certificate from ISO 42001, which covers the management system of the organisation. Sell three agents and you have three AIUC-1 scopes.
Two independent parties do the certifying. An AIUC-accredited auditor examines governance and controls against the requirements, and an accredited testing body attacks the agent itself. AIUC’s certification committee then issues a certificate that runs for 12 months, with technical retesting every quarter. Aegentra plays no part in any of those three steps.
The requirements move every quarter, with releases on 15 January, 15 April, 15 July and 15 October. This page is written to the Q4 2026 release of 15 October 2026, which added 5 requirements and 5 controls, retired 3 requirements and leaves 50 requirements in force. The Q3 2026 release of 15 July 2026 before it updated 8 requirements and 41 controls. A gap assessment that is not dated to a release goes stale without anyone noticing, so ours always carries the release label.
Release assessed
Q4 2026Current release, published 15 October 2026
50Requirements in force across six risk categories
5 + 5Requirements and controls added in Q4 2026; 3 requirements retired
12 monthsCertificate term, with quarterly technical retesting by the accredited testing body
15 January 2027Next scheduled release (Q1 2027)
The six risk categories
A. Data and privacy
How the agent treats personal and customer information: retention, use in training and disclosure.
B. Security
Prompt injection, tool and connector permissions, secrets, access and incident handling.
C. Safety
Harmful outputs, misuse, guardrails and independent adversarial testing.
D. Reliability
Hallucination controls, failure modes, monitoring and fallbacks.
E. Accountability
Ownership, human oversight, logging, change control and customer transparency.
F. Society
Wider harms, fairness and obligations beyond the direct customer.
Counts are as AIUC publishes them in its changelog and Q4 2026 release note, read 30 October 2026. Check the current AIUC-1 standard before relying on a figure.
Who is asking for it
A procurement question, not a New Zealand legal one.
No New Zealand statute names AIUC-1, and neither MBIE’s voluntary responsible-AI guidance nor the Privacy Act 2020 requires it. The pull is commercial. New Zealand companies selling an AI agent into Australian, United States and other enterprise accounts are meeting the question in supplier security questionnaires and procurement checklists, in the same column as SOC 2 and ISO 27001. The buyer is asking whether someone independent has examined the agent’s controls and tried to break it, rather than whether the vendor has written an AI policy.
A second driver arrived on 19 October 2026, when the AICPA and AIUC announced a joint offering that pairs a SOC 2 report with AIUC-1 certification in one CPA-led engagement: the CPA firm performs the SOC 2 and AIUC-1 engagement and issues the report, while AIUC or an accredited technical evaluation body performs the quarterly technical testing. It is being piloted at several CPA firms, with full launch planned for early 2027. A New Zealand vendor that already holds SOC 2 has a short route, and our SOC 2 readiness and AIUC-1 readiness are scoped to fit it.
The local context still matters. Personal information handled by the agent sits under the Privacy Act 2020 and the Office of the Privacy Commissioner’s privacy principles, which is where the data and privacy category of AIUC-1 meets your existing obligations. For the organisational side, our MBIE AI guidance and ISO 42001 planning guide shows how an AI management system can be organised; AIUC-1 then adds the agent-level controls and the mandatory third-party testing that an AIMS does not demand.
Where New Zealand vendors usually start
No certification yet: begin with the agent boundary and the buyer’s exact wording. Readiness builds the governance evidence and the agent-level controls together, and the joint AICPA and AIUC route covers SOC 2 as well if the buyer wants both.
Holding SOC 2: the Trust Services Criteria already cover much of the organisational ground, so readiness concentrates on the agent-specific requirements and the adversarial testing.
Holding ISO 42001: the AIMS supplies most of the governance evidence. Readiness maps your AIMS records to the AIUC-1 requirements through AIUC’s published crosswalk and fills the gaps.
Holding both: a dated gap assessment against the current release, the agent-specific evidence, test preparation and handover. Weeks rather than months for a bounded agent.
AIUC-1 versus ISO 42001
One agent, two standards, different questions.
AIUC publishes an official crosswalk to ISO 42001, together with a reverse mapping from ISO 42001 controls to AIUC-1 requirements, and states that AIUC-1 incorporates most ISO 42001 controls and adds third-party testing. The two sit side by side rather than competing. This table sets out where they part company.
AIUC-1 compared with ISO/IEC 42001:2023
Question
AIUC-1
ISO/IEC 42001:2023
What is certified
One AI agent: its model, tools, data and guardrails
The organisation’s AI management system (AIMS)
Who certifies
AIUC’s certification committee, on the reports of an accredited auditor and an accredited testing body
An accredited certification body
Technical testing
Mandatory adversarial testing of the agent, repeated quarterly
Not required by the standard
Structure
50 requirements in six risk categories, with controls under each
Clauses 4 to 10 plus 38 Annex A reference controls
How often it changes
Quarterly releases
Edition-based; first edition 2023
Term
12 months, with quarterly retesting
Three-year cycle with annual surveillance
Who asks for it
Enterprise, Australian and US buyers of AI agents
Procurement, regulators and boards governing AI use
What carries across
AIMS governance evidence covers much of the accountability and data categories
An AIUC-1 evidence pack supplies agent-level operating evidence for the AIMS
AIUC also publishes crosswalks to the NIST AI RMF, the EU AI Act, OWASP, MITRE ATLAS and SOC 2. If you hold ISO 42001 or are building towards it with our ISO 42001 implementation service, the crosswalk is where readiness begins.
Aegentra does not certify against AIUC-1, does not audit for AIUC, does not run the accredited adversarial testing and holds no AIUC accreditation of any kind. Only AIUC accredits auditors and testing bodies, and only AIUC issues certificates. You appoint the accredited auditor; we work to that auditor’s information requests and hand over the readiness outputs. Readiness support does not guarantee a certificate, a certificate date or a customer contract.
The firms below are the AIUC-accredited auditors on AIUC’s register as read on 30 October 2026. They are independent third parties with no relationship to Aegentra. Accreditation follows a witness audit, so a provisional firm is completing that step. The register changes, so confirm it on AIUC’s accredited-auditors page before you appoint anyone.
AIUC-accredited auditors, as listed by AIUC on 30 October 2026
Firm
Status
Accreditation period from
Schellman
Fully accredited
1 November 2025
Coalfire
Provisional
1 May 2026
A-LIGN
Provisional
1 August 2026
BDO
Provisional
Listed as provisional
Grant Thornton
Provisional
Listed as provisional
Mastermind
Provisional
Listed as provisional
Sensiba
Provisional
Listed as provisional
Who does what
Responsibilities from readiness to the certificate
Stage
Who
What happens
Readiness
Aegentra
Gap assessment, policies and ownership records, evidence pack, crosswalk, test-readiness briefing and handover.
Controls
Your organisation
Owns and operates the agent and its controls, and completes the agreed actions.
Audit
AIUC-accredited auditor
Examines governance and controls against the requirements and reports to AIUC. Appointed by you, outside Aegentra.
Testing
Accredited testing body
Adversarial technical testing of the agent, then quarterly retests. Outside Aegentra.
Certificate
AIUC certification committee
Decides, and issues the certificate for 12 months.
Decision
Your customer
Decides whether the certificate answers the question in its questionnaire.
What the engagement can produce
Useful outputs, agreed in advance.
A written scope naming the agent, the accountable lead, assumptions and exclusions
A requirement-by-requirement gap list with owners and priorities, dated to the release
A requirement-to-control-to-evidence map built in the order an auditor samples it
Readiness outputs handed to your appointed accredited auditor, with the opening meeting attended where the scope includes it
Starting prices and indicative ranges are for the stated scope, not an unconditional fixed fee. Other services are priced by scoped quote. Academy course prices are separate from consulting, audit and technical-service fees.
Who delivers the work
Implementation and governance capability
Our consultants bring experience across management systems, cybersecurity governance, enterprise risk and internal controls. We select the relevant implementation and specialist capability for the system you need to build.
Qualifications, credentials and training across our team
ISO/IEC 42001 Lead Implementer training
ISO/IEC 42001 Lead Auditor
ISO/IEC 27001 Lead Implementer
CISA — Certified Information Systems Auditor
CISSP — Certified Information Systems Security Professional
Qualifications, professional credentials and formal training are held across our delivery team, not by every person. We match the relevant capability to your scope and confirm the assigned people, responsibilities and current qualification evidence before work begins.
Our team supports readiness only. Aegentra holds no AIUC accreditation, is not an AIUC-accredited auditor or testing body, and does not issue AIUC-1 certificates. The ISO 42001 and security credentials listed support the crosswalk and evidence work; they are not an AIUC qualification. The accredited auditor you appoint, the accredited testing body and AIUC’s certification committee retain their responsibilities.
Decide exactly what is being certified: which product, model, tools, connectors, data and guardrails, and which buyer question it has to answer. The written scope names the accountable lead and records assumptions and exclusions. One agent, one scope; a shared platform is recorded once and referenced.
Assess against the Q4 2026 release
Walk the agent and the organisation through each requirement and its controls, reuse the evidence you already hold through AIUC’s published crosswalks, and record a dated gap list with an owner and a priority against every line, so the next quarterly release is a delta rather than a restart.
Build and evidence
Policies, ownership records and procedures are drafted or tailored inside your own systems. Your team operates the controls; we organise the evidence in the order an auditor samples it and brief the team on what the accredited testing body will try. Engineering changes to the agent stay with your engineers or a separately scoped service.
Hand over
We pass the readiness outputs to the accredited auditor you appoint and attend the opening meeting where the scope includes it. The auditor examines, the testing body tests and AIUC’s certification committee decides. None of that is Aegentra’s to promise, and findings are represented as they are.
We deliver online and arrange visits by appointment where agreed. Scope, assigned competence, timing, access requirements and any travel are confirmed before an engagement.
An illustrative starting point
What a useful first decision looks like.
Fictional example to explain the service, not a New Zealand client case study or a promised outcome.
A Wellington software company sells a customer-support agent to an Australian insurer. The insurer’s security questionnaire asks for AIUC-1 next to SOC 2, which the vendor already holds. The first useful decision is not to buy a policy template: it is to fix the agent boundary, map the existing SOC 2 evidence through AIUC’s crosswalk, and find out whether the insurer will accept the joint SOC 2 and AIUC-1 route its CPA firm can now offer.
Questions for the first discussion
Which agent, and which buyer asked for AIUC-1 in which words?
What does the agent reach: model, tools, connectors, customer data and guardrails?
Which certifications do you already hold, and when were they last examined?
Before you begin
Your questions, answered.
Who chooses the accredited auditor?+
You do. We explain AIUC’s register and what fully accredited and provisional mean, but the appointment is yours and the firms on the register are independent third parties with no relationship to Aegentra. Once appointed, we work to that auditor’s information requests and hand over the readiness outputs; the audit itself, the testing and the certificate decision stay outside Aegentra.
What can be included in AIUC-1 readiness?+
The agreed scope can include the agent boundary: product, model, tools and connectors, data, guardrails and the buyer question it must answer; gap assessment against every requirement of the Q4 2026 release, dated to that release; policies, ownership records and the evidence pack an accredited auditor samples; crosswalk of existing ISO 42001, SOC 2 or ISO 27001 evidence, test readiness and auditor handover. We confirm the exact deliverables and exclusions before work begins, rather than treating every organisation as the same project.
How much does this service cost?+
AIUC-1 readiness starting package: From NZ$8,125. Per agent, for one AI agent with a single clearly bounded purpose, one legal entity, existing core security controls and an internal owner able to complete agreed actions. Two agents on a shared platform are quoted together after scoping; a wider or less bounded scope needs a wider quote. Prices in NZD, excluding any applicable tax. Your written proposal confirms the scope, tax treatment and total payable before you approve the work.
How do you deliver the work in New Zealand?+
We deliver online and arrange visits by appointment where agreed. Scope, assigned competence, timing, access requirements and any travel are confirmed before an engagement.
What should we provide before a proposal?+
Bring answers to these starting questions: Which agent, and which buyer asked for AIUC-1 in which words? What does the agent reach: model, tools, connectors, customer data and guardrails? Which certifications do you already hold, and when were they last examined? A short initial discussion should establish the boundary and next evidence needed. Do not email secrets or detailed personal records before secure handling has been agreed.
What is AIUC-1, and who issues the certificate?+
AIUC-1 is a certification standard for one specific AI agent, published by the Artificial Intelligence Underwriting Company in July 2025 and revised every quarter. Its requirements sit in six risk categories: data and privacy, security, safety, reliability, accountability and society. An AIUC-accredited auditor examines governance and controls, an accredited testing body runs adversarial testing against the agent, and AIUC’s certification committee issues the certificate, which lasts 12 months with quarterly retesting.
Does Aegentra certify, audit or test against AIUC-1?+
No. Aegentra provides readiness only. We do not certify, do not audit for AIUC, do not run the accredited adversarial testing and hold no AIUC accreditation. You appoint an accredited auditor from AIUC’s own register, which lists Schellman as fully accredited and Coalfire, A-LIGN, BDO, Grant Thornton, Mastermind and Sensiba as provisional; we hand that auditor the readiness outputs and can attend the opening meeting where the scope includes it.
What does AIUC-1 readiness cost in New Zealand?+
Aegentra’s starting package is from NZ$8,125 per agent for the bounded scope set out in the pricing section, with the final fixed fee confirmed after a scoping call. Prices are in NZD before any applicable tax; tax follows the billing address on the invoice, and a New Zealand billing address adds nothing. The accredited audit, the adversarial technical testing, the certificate and any engineering changes to the agent are separate costs charged by other parties or quoted separately.
What does “per agent” mean?+
AIUC-1 certifies an agent, not a company, so each distinct agent with its own purpose, model, tools and data boundary is its own scope and its own certificate. Two agents that share a platform are quoted together after scoping, because much of the governance evidence is common to both. A business with three unrelated agents has three scopes.
How long does AIUC-1 readiness take for a New Zealand vendor?+
It turns on how bounded the agent is, what evidence you already hold and which release is in force. For a bounded agent with existing security controls, readiness is normally a matter of weeks rather than months. Audit and testing dates belong to the accredited auditor and the testing body, so we do not promise a certificate date, and the work runs online across New Zealand with visits by appointment where agreed.
We already hold ISO 42001 or SOC 2. How much carries across?+
A good deal of the governance evidence, once its scope is checked against the agent. AIUC publishes crosswalks to ISO 42001, SOC 2 and the NIST AI RMF, and we map your existing evidence through them rather than rebuilding it. AIUC-1 then adds agent-specific controls and mandatory adversarial testing that neither standard supplies. A SOC 2 holder can also use the joint SOC 2 plus AIUC-1 engagement the AICPA and AIUC announced on 19 October 2026.
Which AIUC-1 release do you assess against, and what happens when it changes?+
The release named on this page, currently Q4 2026, published 15 October 2026. AIUC-1 changes every quarter, on 15 January, 15 April, 15 July and 15 October, and the next release is due on 15 January 2027. Our gap assessment is dated to a release, and the handover notes which requirements changed since, so a new release becomes a delta to work through rather than a fresh start.
Do New Zealand companies need AIUC-1?+
No New Zealand law requires it, and neither MBIE’s responsible-AI guidance nor the Privacy Act 2020 mentions it. The demand comes from Australian, US and other enterprise buyers writing it into procurement and security questionnaires. If your agent sells into those buyers, readiness costs less before the questionnaire arrives than after a deal stalls on it. If it does not, ISO 42001 may be the better first step, and we will tell you so.
Start with the requirement, the systems involved and the outcome you need. Please do not send passwords, sensitive records or unrestricted tenant access in an initial enquiry.