Key Takeaways:
- The AICPA and AIUC now offer a combined SOC 2 and AIUC-1 report through a single CPA-led engagement.
- SOC 2 covers organization-wide security and data controls, while AIUC-1 adds agent-specific controls and live behavioral testing.
- Quarterly retesting and quarterly standard updates keep AIUC-1 current as agents and attack techniques change.
- One auditor and shared evidence make the combined report more rigorous without doubling the compliance workload.
The AICPA has formally stepped into AI assurance. In October 2026, it partnered with the Artificial Intelligence Underwriting Company (AIUC) to launch a combined SOC 2 and AIUC-1 offering, pairing one of the most widely requested security attestations with what AIUC calls the world’s first standard for AI agent safety, security, and reliability.
360 Advanced can now support this offering, which is delivered through one CPA-led engagement and results in one report. For organizations building or deploying AI agents, the combination produces a more rigorous picture of risk than either framework provides on its own.
WHERE SOC 2 ALONE LEAVES A GAP
SOC 2 remains the foundation for data and security assurance. A licensed CPA examines controls relevant to security, availability, processing integrity, confidentiality, or privacy and issues an opinion on their design and operating effectiveness. Access management, encryption, change management, and incident response all sit squarely in scope.
AI agents introduce risks that a traditional controls examination was never designed to measure. Agents access data, call tools, and execute transactions on their own. They are non-deterministic, so the same input can produce different actions. They face new attack surfaces, including prompt injection, jailbreaks, and hallucinated outputs. And their underlying models, prompts, and tools change constantly, which means an annual point-in-time review can go stale quickly.
WHAT AIUC-1 ADDS TO A STANDARD SOC 2 REPORT
AIUC-1 is a standard built specifically for AI agent security, safety, and reliability. Its controls span six principles: data and privacy, security, safety, reliability, accountability, and society. They map to MITRE ATLAS, ISO/IEC 42001, and the EU AI Act. When paired with SOC 2, it adds three layers of rigor:
- Live behavioral testing. AIUC-1 evaluates how agents behave across thousands of risk and attack scenarios, such as hallucinations, unsafe tool calls, and harmful outputs. According to AIUC, the highest-severity findings must be remediated before certification.
- Agent-specific controls. Requirements address risks SOC 2 criteria don’t name directly, including agent permissions, output safety, and accountability for agent actions.
- Continuous validation. AIUC-1 requires quarterly technical retesting, and the standard itself is updated quarterly as new agent risks and attack techniques emerge.
The result is a report that covers the organization’s security foundation and validates the behavior of the agents running on top of it. As AIUC cofounder Rajiv Dattani put it in the announcement, “What sets AIUC-1 apart is live testing of AI agent behavior, not just a review of policies and guardrails.”
HOW THE SCOPE DIVIDES
Each framework covers the domain it was built for. A SOC 2 examination addresses organization-wide systems and data controls. AIUC-1 focuses on the agents in scope, either a specific deployed agent or representative agents built on a platform.
Because both frameworks sit inside one engagement, organizations work with one audit team and one evidence process. Overlapping evidence can be reused across both, and crosswalks help map controls between the frameworks. AIUC estimates the combined engagement takes six to 10 weeks and is designed to follow an organization’s existing compliance cycle to limit the workload on GRC teams.
WHO SHOULD PAY ATTENTION
The combined report is most relevant for organizations that:
- Sell AI agents or agentic features to enterprise customers
- Already maintain a SOC 2 report and field AI questions during security reviews
- Face board or customer pressure to show independent proof of AI safety
- Deploy internal agents with access to sensitive data or financial workflows
HOW COMPANIES CAN PREPARE NOW
Even before full launch, organizations can position themselves for a combined engagement. Start by inventorying the agents you build or deploy and deciding which carry the most business risk. Review your current SOC 2 controls for places where AI already touches access, change management, and vendor oversight. Plan for technical testing access to your agents, and set aside remediation capacity, since behavioral testing often surfaces issues that policy reviews miss.
Both SOC 2 reports and AIUC-1 certifications renew annually, so aligning the two cycles early will pay off in every year that follows.
THE BIGGER PICTURE
AI adoption is moving faster than most governance programs. A combined SOC 2 and AIUC-1 report gives organizations a way to use AI with confidence while keeping independent, CPA-led oversight in place. That combination of strong controls and tested agent behavior is quickly becoming what enterprise buyers expect.
360 Advanced is a people-first, technology-forward cybersecurity and compliance firm and a licensed CPA firm. Our team helps organizations strengthen the SOC 2 foundation that AI assurance builds on. Learn more about how we support this new combined offering as well as more about AIUC-1 and AI agent risk. People First. Technology-Forward.
