Skip to content
Atesto
Create account
All articles

Audits and evidence

AICPA proposes 2027 SOC 2 updates for AI and emerging tech

The AICPA has proposed updates to the SOC 2 Trust Services Criteria, adding specific controls for AI governance, data poisoning and automated decisions.

The American Institute of CPAs proposed a significant update to the SOC 2 Trust Services Criteria on October 28, 2026. The exposure drafts (opens in a new tab) issued by the AICPA Assurance Services Executive Committee aim to modernize the framework that independent auditors use to evaluate information system controls. This proposal introduces explicit considerations for artificial intelligence and other emerging technologies. For companies that rely on SOC 2 reports to prove their security posture to customers, the update signals a shift in how evidence will be collected and evaluated.

Service organizations use SOC 2 examinations to demonstrate that their controls over security, availability, processing integrity, confidentiality and privacy meet established industry standards. As companies integrate new technological models into their service delivery, the risks associated with data processing and third-party dependencies have changed. The AICPA press release (opens in a new tab) notes that the criteria must adapt as emerging technologies "alter third-party risk considerations." Amy Pawlicki, the AICPA vice president of assurance and advisory innovation, stated that the way organizations manage information and identify risk has changed substantially over the past decade.

To address this shift, the proposed 2027 update restructures the criteria. It removes duplicative concepts and adds new points of focus specifically tailored to advanced technologies. This moves the audit process toward a more concise and uniform understanding of the intention behind each criterion. Instead of relying on older interpretations of legacy controls, the proposal establishes specific points of focus for how artificial intelligence must be governed, monitored and secured within a service organization.

Who it applies to

The proposed updates directly affect service organizations that undergo SOC 2 examinations.

The framework also applies to the licensed, independent CPAs who perform the attestation engagements under the AICPA Attestation Standards. Auditors will use the revised criteria to evaluate whether a service organization has designed and operating controls that effectively mitigate the risks introduced by emerging technologies.

Finally, the changes impact the report users. The updated reports are intended to provide these users with a clearer, more uniform understanding of how their vendors manage artificial intelligence and automated systems.

What it requires

The exposure drafts introduce targeted restructuring to improve the clarity and practical application of the Trust Services Criteria and the Description Criteria. The proposal consolidates overlapping concepts and revises existing language to ensure a concise understanding of the intention behind each criterion. Where greater emphasis is needed, the AICPA has added entirely new criteria.

The most prominent addition is the expanded focus on artificial intelligence. The proposal requires organizations and auditors to evaluate controls against specific risks introduced by these advanced models. The new points of focus address artificial intelligence governance, ensuring that companies have formal oversight over how these systems are deployed and maintained.

The requirements extend to the technical operation of the models. Organizations will need to demonstrate controls around model monitoring to ensure systems function as intended over time. The criteria also introduce specific focus areas for prompt manipulation and data poisoning. This means service providers must implement safeguards that prevent malicious inputs from altering the behavior of the model or corrupting the training data.

Transparency of artificial intelligence outputs is another key requirement. Companies must be able to show how they manage training data and how they communicate the nature of automated outputs to their users. Furthermore, the proposal addresses the risks associated with automated decision-making. Service organizations that rely on algorithms to make operational or security decisions will need to provide evidence showing how those decisions are controlled, reviewed and secured.

By introducing these specific points of focus, the AICPA aims to ensure the criteria remain relevant. Auditors will look for evidence that controls effectively address these new categories of risk.

The dates

The AICPA Assurance Services Executive Committee is currently seeking public comment on the exposure drafts. The consultation period remains open through January 15, 2027.

If the proposal is approved next year, the 2027 SOC 2 Trust Services Criteria will become effective for attestation engagements beginning on or after December 31, 2027. The AICPA has stated that early adoption of the new criteria will be permitted for organizations that wish to align their audits with the updated framework before the mandatory effective date.

What to check now

Companies preparing for future SOC 2 renewals can take practical steps to assess their readiness for the proposed changes.

  • Review the exposure drafts published by the AICPA to understand the specific language and new points of focus.
  • Inventory all artificial intelligence tools and automated decision systems currently used within the service delivery environment.
  • Map existing security controls against the proposed requirements for model monitoring, prompt manipulation and data poisoning.
  • Assess how training data is managed and documented, ensuring there is a clear record of provenance and security.
  • Evaluate third-party dependencies to determine if vendors providing artificial intelligence capabilities meet the proposed governance and transparency standards.
  • Discuss the proposed 2027 criteria with internal compliance teams and external auditors to plan for potential early adoption or the eventual transition.

The modernization of the SOC 2 framework reflects the growing necessity to secure advanced technologies. By moving toward explicit points of focus for artificial intelligence, the attestation process will ask service organizations to show exactly how they govern and monitor these models.

Sources

  1. AICPA Exposure Drafts (opens in a new tab) · www.aicpa-cima.com
  2. AICPA Press Release (opens in a new tab) · www.aicpa-cima.com

This article is general information, not legal advice. For your company's situation, talk to a qualified lawyer.

This article was written with AI and checked, sentence by sentence, against the sources linked above.