SOC 2 + AIUC-1: What the AICPA and AIUC Partnership Means for Your Next Audit

On October 19, 2026, the American Institute of Certified Public Accountants (AICPA) and the Artificial Intelligence Underwriting Company (AIUC) announced a joint offering that pairs a SOC 2 examination with AIUC-1 certification in a single CPA-led engagement.

For organizations building or deploying AI agents, the offering adds a standard written specifically for agent security, safety, and reliability to a CPA attestation. The offering is in a pilot period, and the partners expect a full launch in early 2027.

What the AICPA and AIUC partnership is

According to the AICPA’s announcement, more than 40,000 organizations obtained a SOC 2 attestation or an AIUC-1 certification last year. The new offering lets an organization earn both through one engagement, with one auditor, resulting in one report. It does not replace your SOC 2. The engagement produces a full SOC 2 attestation alongside the AIUC-1 certification, and your SOC 2 stays on its annual cycle.

In an October 27 session with the AIUC-1 consortium, the AICPA explained that rather than create an AI framework of its own, it concluded AIUC-1 has the elements CPAs need to provide assurance over AI agents. AIUC-1 was developed with a consortium of more than 250 CISOs and security leaders and is updated every quarter to keep pace with how agents and attacks change.

The two frameworks remain independent. AIUC-1 will continue its quarterly update cycle, and SOC 2 will continue on its own cycle. Several CPA firms are delivering the joint SOC 2 + AIUC-1 engagement now, and the partners expect a full launch in early 2027 once initial engagements conclude.

On October 28, 2026, the AICPA issued an exposure draft updating the SOC 2 Trust Services Criteria, including expanded considerations for emerging technologies such as AI. If approved, the updated criteria would apply to engagements beginning on or after December 31, 2027, with early adoption permitted, and public comments are due January 15, 2027.

AI agents now hold credentials, call APIs, and act inside core systems, and enterprise buyers are sending AI security questionnaires that can slow procurement. An independent report covering both the system and the agent gives buyers and sellers a shared reference point.

How the combined engagement works

The details below draw on AIUC’s October 27 consortium briefing and its public documentation. The program is still in its pilot period, so confirm reporting structure, criteria, and certification maintenance against current program documentation before relying on them.

One engagement, one auditor, one report

A joint engagement follows one audit process with one auditor, and controls that overlap between the two frameworks are tested once. The result is two distinct outcomes delivered together: a SOC 2 attestation, which is a CPA’s opinion issued under the AICPA’s attestation standards, and an AIUC-1 certification, which AIUC issues under its own program.

AIUC’s published SOC 2 crosswalk maps AIUC-1 requirements to all five Trust Services Criteria categories, and the organization still decides whether its SOC 2 covers Security alone or additional categories.

AIUC described the combined report in four parts:

• Introduction: the auditor’s attestation, management’s assertion, and the scope of the audit 
• SOC 2 attestation: the full report on the selected Trust Services Criteria 
• AIUC-1 certification: the full certification report on applicable AIUC-1 requirements 
• Technical evaluation results: the results of testing agent behavior  

This is a deliberate departure from a typical “SOC 2+” report. An auditor cannot add a mapping to AIUC-1 in an additional section of the SOC 2 report (Section 5) and claim certification, because the technical evaluation of the agent itself is central to the standard. AIUC-1 remains an open standard, so organizations can implement its controls ahead of pursuing certification.

Scope sits at two levels

AIUC’s public scoping guidance limits AIUC-1 to agentic AI systems. In its October 27 briefing, AIUC described SOC 2 as covering the organization’s systems in scope and AIUC-1 as covering the agents within them, and outlined two ways it certifies. An agent platform is tested through representative agents built on it, with a control focus on core security, secure defaults, and documentation. An individual agent is tested directly, with a focus on that agent’s configuration plus core security. In both cases, AIUC described the underlying model as outside AIUC-1’s scope.

Timing: a period for SOC 2, a point in time for AIUC-1

According to AIUC, all SOC 2 and AIUC-1 controls are tested annually in one audit. A SOC 2 Type 2 assesses control effectiveness over a period of time. AIUC’s public FAQ describes AIUC-1 certification as a point-in-time assessment, valid for 12 months, with agent evaluations required at least every three months to keep it valid.

If a quarterly evaluation is not passed, the AIUC-1 certification goes stale, and the SOC 2 attestation is not affected. AIUC noted this may change as the frameworks integrate further, and that any such change would apply only to future reports.

The evaluations and the process

In the same briefing, AIUC described evaluations that run more than 2,000 risk and attack scenarios against the agent, drawn from a risk taxonomy, an attack taxonomy, and a database of real-world incidents. Results are graded on a P0 to P4 severity scale, and any P0 or P1 finding must be remediated before certification. AIUC outlined the engagement in four phases:

AIUC also noted two boundaries. AIUC-1 currently has no link to SOC 1 or financial reporting reliance. And AIUC positions the offering for organizations that develop agents sold to enterprises, deploy agents in high-risk environments, face agent-related regulatory requirements, use coding agents for production code, or field AI security reviews from enterprise buyers.

Open questions before the 2027 launch

The design of SOC 2 + AIUC-1 is settled. Operational details are still being worked out in early engagements. The questions below are the ones security, compliance, and procurement teams are likely to raise as they plan budgets and audit calendars, and published guidance from the AICPA and AIUC may address them before the full launch.

Review period and timing

• Agents that go live partway through a period. If an agent enters production midway through a SOC 2 Type 2 review period, can it still be included, given that AIUC-1 is assessed at a point in time? How should the system description reflect it? 
• Where the point-in-time date falls. Is the AIUC-1 assessment date aligned to the end of the SOC 2 review period, or can it be set independently? 
• How quarterly evaluations relate to the annual report. Does the CPA review quarterly evaluation results that fall inside the review period, or does AIUC manage them entirely between audits? 
• Bridge letters. How should a bridge letter speak to AIUC-1 status when a quarterly evaluation lands between report dates? 
• First-year options. Will a SOC 2 Type 1 + AIUC-1 path be available for organizations that are new to SOC 2? 

Scoping 

• Agents and the SOC 2 system boundary. Must every in-scope agent sit inside the SOC 2 system boundary, or can an organization certify an agent that its existing SOC 2 does not describe? 
• Foundation model providers. With the model outside AIUC-1’s scope, how should a model provider be reflected in the SOC 2 portion? Depending on the system boundary and the service relationship, it may be a subservice organization presented under the carve-out or inclusive method, or treated another way. 
• Platform certifications. When a platform is certified through representative agents, how should that platform’s own customers interpret the certification for agents they build and configure themselves? 
• Standard versions. With AIUC-1 updated quarterly, which version applies to a given engagement, and how much lead time will organizations have to adopt new requirements? 

Management assertion and system description 

• What management asserts. Is there one management assertion covering both SOC 2 and AIUC-1, or a separate assertion for each? What does management assert about the AIUC-1 requirements? 
• Describing an agent. What should a system description include for an agentic system, such as models, tools, guardrails, and evaluation approach? Will the AICPA issue guidance on this before the full launch? 

Reporting and distribution 

• Detail in the evaluation results. How much of the technical evaluation detail appears in the report, and how are P2 through P4 findings presented? 
• Restricted use and a public certificate. SOC 2 reports are restricted-use documents, while the AIUC-1 certification is designed to be shown publicly. How should organizations share each with prospects and customers? 
• Current status for relying parties. Because certification depends on quarterly evaluations, how can a customer confirm that a vendor’s AIUC-1 certification is current on the day they review it? 

Looking ahead 

• The updated Trust Services Criteria. As the proposed Trust Services Criteria update adds considerations for AI, how will it relate to AIUC-1, and will the overlap reduce duplicated testing? 
• Deployers relying on vendors. Can an organization that deploys a vendor’s certified agent rely on that certification in its own SOC 2, and what complementary user entity controls would apply? 

How to prepare now 

Much of the groundwork can start before these questions are resolved. 

1. Inventory your agents. List the agents you build or deploy, what data and tools each one can reach, and whether each sits inside your current SOC 2 system boundary. 
2. Decide whether you are certifying a platform or an agent. The answer shapes testing, control focus, and how your own customers will read the result. 
3. Read the AIUC-1 standard and the SOC 2 crosswalk. Both are publicly available, and the crosswalk shows where your existing SOC 2 controls already carry weight and where agent-specific work begins. 
4. Line up your audit calendar. Map your current SOC 2 review period against a likely AIUC-1 assessment date and the quarterly evaluation cycle, so certification and attestation land when your buyers need them. 
5. Bring your questions to your auditor early. Scoping decisions made in the first two weeks set the cost and the timeline for everything that follows. 

Where ISO 42001 fits 

Many organizations already hold, or are pursuing, ISO 42001 certification. ISO 42001 establishes an AI management system that runs across the organization, including how AI objectives are set, how AI risk and impact are assessed, and how AI use is monitored and improved. SOC 2 + AIUC-1 provides assurance over a defined system and the agents inside it.

AIUC publishes a crosswalk from AIUC-1 to ISO 42001, so work done for one may provide reusable evidence for the other, subject to scope and each scheme’s own requirements. An organization with a working AI management system often already has many of the policies and review processes AIUC-1 looks for.

Frequently asked questions

Does SOC 2 + AIUC-1 replace our existing SOC 2?

No. The engagement produces a full SOC 2 attestation alongside the AIUC-1 certification. Your SOC 2 continues on its annual cycle, and the AIUC-1 component adds agent-specific requirements and technical evaluation.

If an agent fails a quarterly evaluation, does our SOC 2 report fail too?

No. AIUC has told its consortium that the two outcomes are independent. A failed quarterly evaluation causes the AIUC-1 certification to go stale, and the SOC 2 attestation is unaffected.

Can our auditor map our SOC 2 controls to AIUC-1 in Section 5 and call it certified?

No. Certification requires the full joint engagement, including AIUC’s technical evaluation of the agent.

Does AIUC-1 cover the underlying AI model?

No. According to AIUC’s October 27 briefing, AIUC-1 certifies agent platforms and individual agents, and the model is outside its scope.

Does this help with SOC 1 or financial reporting?

Not today. AIUC has said AIUC-1 currently has no link to SOC 1.

How long does a joint engagement take?

AIUC’s October 27 briefing outlines roughly 1 to 2 weeks for scoping, 2 to 6 weeks for evidence collection with evaluations running alongside it, and 1 to 2 weeks to complete the audit. Your own timeline will depend on scope and readiness.

When is the full launch?

The AICPA expects the full launch in early 2027, once the initial engagements conclude.

A-LIGN brings more than 20 years of SOC, ISO, and federal auditing experience to AIUC-1 audits, in partnership with AIUC. If you are deciding how agents fit into your next SOC 2, or whether to pursue AIUC-1 first, contact us to talk through scope and timing.