SOC 2 controls list

There is no official, AICPA-issued SOC 2 controls list. The AICPA publishes trust services criteria and points of focus, and your organisation designs the controls that meet them. What follows is the next best thing: the controls auditors commonly test in practice, organised by criteria family, each with its typical owner, frequency, evidence artefact and the exception it most often maps to.

Criteria familyWorking controls covered hereWhere first-timers trip
Governance and people (CC1, CC2)Roles, policies, oversight, screening, training, joiner and leaver processPaper policies with no evidence of operation
Risk and monitoring (CC3, CC4, CC5)Risk register, fraud and change assessment, internal review, corrective actions, controls matrixStale register, findings never closed
Access (CC6)MFA, unique accounts, provisioning, access reviews, encryption, malware protection, physical accessLeavers retaining access, rubber-stamped reviews
System operations (CC7)Vulnerability management, logging, incident response and testingScans run, findings never remediated
Change management (CC8)Code review, environment separation, emergency change, hardening standardsSelf-approved releases
Risk mitigation (CC9)Vendor reviews, vendor SOC 2 report collection, continuity planning, recovery testingBackups that are never restore-tested
Availability (A-series, if in scope)Uptime and capacity monitoring, backup monitoring, restore testsNo evidence anyone watches the monitors
Processing integrity (PI-series, if in scope)Input validation, reconciliations, failed-job handlingFailed jobs nobody reviews
Confidentiality (C-series, if in scope)Classification, NDAs, retention and disposalA retention schedule with no deletion evidence
Privacy (P-series, if in scope)Privacy notice, consent records, request handling, breach notificationA notice that no longer matches actual data use

Why listen to us

Atoro is an Irish AI governance and cyber compliance consultancy for software companies, and the first consultancy in Europe certified against ISO/IEC 42001 (certificate AIMS-AT-120224, issued by A-LIGN under ANAB accreditation). Atoro states it has delivered more than 200 certifications (atoro.io, checked 08-08-2026). We prepare software companies for SOC 2 examinations, so the catalogue below reflects what auditors actually ask for, not what a template says they should.

One paragraph of orientation, then this page stays at control level

The AICPA does not publish a controls list. It publishes the Trust Services Criteria, and points of focus that “represent important characteristics of the criteria”; the document states plainly that using the criteria “does not require an assessment of whether each point of focus is addressed” (AICPA, TSP section 100, par. .03-.04, 2017 edition with revised points of focus 2022, accessed 19-08-2026). The five categories, the CC1 to CC9 criteria structure and the full argument against the idea of an official checklist are covered in our companion guide, SOC 2 requirements, so we will not repeat them here. This page works one level down: the controls companies actually implement, which are practitioner practice, not AICPA requirements. Where we describe what auditors expect, that is market practice observed across SOC 2 examinations and audit-prep work, not an AICPA rule.

How to read this catalogue

Each control below carries four fields. Owner: the role that usually holds the control in a software company; assign a named person, not a team. Frequency: how often the control normally runs. Frequencies are common practice, not AICPA-set rules; your control design sets the cadence and the auditor tests whether you kept to it. Evidence artefact: the record an auditor will ask to see. Common exception: the finding this control most often produces, drawn from exception patterns auditors report in the market (per the SERP research record for this cluster: soc2auditors.org’s evidence and gap columns and Scrut’s exception-risk triage, checked 19-08-2026). If you take one thing from this page, take the evidence column: a control without an artefact is a claim, not a control.

Governance and people controls (CC1 and CC2)

ControlTypical ownerTypical frequencyEvidence artefactCommon exception
Defined security roles and reporting linesLeadershipReviewed annuallyOrg chart and role descriptionsRoles exist on paper, no evidence of oversight
Approved policy suite (security, acceptable use, data handling)Security leadAnnual review and sign-offVersioned policies with approval recordPolicies unsigned or stale
Leadership or board review of securityLeadershipAt least annualMeeting minutes with actionsReviews minuted but actions never tracked
Background checks for new hiresHRPer hireScreening recordsChecks missing for early employees
Security awareness trainingHR or IT leadAt hire, then annuallyCompletion recordsTraining runs, completion never enforced
Onboarding and offboarding checklistsHR and IT leadPer joiner and leaverCompleted checklists tied to ticketsLeavers processed late

Risk and monitoring controls (CC3, CC4 and CC5)

ControlTypical ownerTypical frequencyEvidence artefactCommon exception
Risk register tied to commitments and system requirementsSecurity leadReviewed at least annually and on significant change (common practice)Risk register with scoring and ownersRegister stale or generic
Fraud risk considered in the assessmentLeadershipWith each assessmentFraud section in the risk registerFraud never discussed
Change impact assessmentSecurity leadPer major change (new product, new region, acquisition)Change review notesAssessment skipped during fast growth
Internal audit or structured self-assessmentSecurity leadAnnualInternal review report and findings logFindings raised, never closed
Corrective action trackingControl ownersPer findingAction log with closure evidenceThe same exceptions recur year on year
Controls matrix mapping each control to the criteria it addressesSecurity leadUpdated whenever scope changesThe matrix itselfControls mapped to criteria they do not actually address

Access controls (CC6)

ControlTypical ownerTypical frequencyEvidence artefactCommon exception
Multi-factor authentication on remote and administrative accessIT leadContinuousIdentity provider policy export, MFA coverage reportGaps on service accounts and legacy apps
Unique user accounts, no shared credentialsIT leadContinuousAccount listingShared admin accounts survive from the early days
Joiner, mover and leaver provisioning tied to HR eventsIT lead and HRPer eventTickets linked to HR recordsLeaver access removed days or weeks late
Periodic access reviewsIT lead and system ownersQuarterly (common practice)Signed review recordsReviews rubber-stamped without checking
Encryption of data at rest and in transitEngineering leadContinuousKey management and TLS configuration evidenceLegacy data stores left unencrypted
Malware prevention and detection on endpoints (CC6.8)IT leadContinuousEndpoint protection deployment and alert reportsCoverage gaps on contractor devices
Restricted physical access to sensitive locations (CC6.4)Office lead; cloud provider relianceContinuousBadge logs; data centre assurance reports from the cloud providerBadge access lists never reviewed

System operations controls (CC7)

ControlTypical ownerTypical frequencyEvidence artefactCommon exception
Vulnerability scanning with remediation trackingSecurity leadMonthly or quarterly (common practice)Scan reports plus remediation ticketsScans run on schedule, findings never fixed
Centralised logging and alertingEngineering leadContinuousLog retention configuration, sample alertsLogs exist, nobody reviews the alerts
Incident response plan with defined rolesSecurity leadReviewed annuallyIncident response planPlan written, never tested
Tabletop or simulated incident exerciseSecurity leadAnnual (common practice)Exercise notes and resulting actionsFirst test happens inside the audit window
Security event triage through to incident determinationSecurity leadPer eventTriage and investigation recordsEvents closed without analysis

Change management controls (CC8)

ControlTypical ownerTypical frequencyEvidence artefactCommon exception
Code review and approval before mergeEngineering leadPer changePull request approvalsSelf-approved releases
Separation of development and production environmentsEngineering leadContinuousEnvironment and access configurationDevelopers deploying straight to production
Emergency change processEngineering leadPer eventEmergency change record with retrospective approvalEmergencies used to bypass review
Documented hardening or baseline configuration standardEngineering leadReviewed annuallyHardening standard plus baseline configurationThe standard exists in someone’s head, not in a document

Vendor and continuity controls (CC9)

ControlTypical ownerTypical frequencyEvidence artefactCommon exception
Vendor risk assessment at onboardingSecurity lead or procurementPer vendorVendor review recordsVendors approved with no review at all
Collection and review of critical vendors’ SOC 2 reports, including any complementary user entity controlsSecurity leadAnnually per critical vendor (common practice)Vendor reports plus review notesReports collected, never read
Business continuity and disaster recovery planEngineering or operations leadReviewed annuallyBCP and DR documentPlan copied from a template, never adapted
Recovery testingEngineering leadAnnual test (common practice)Restore test or DR exercise reportBackups run for years, restores never tested

Controls for the optional categories

If availability, processing integrity, confidentiality or privacy sit inside your scope, auditors commonly test controls like these alongside the security controls above. Scoping itself is covered in SOC 2 requirements; this is the working layer only.

CategoryControlTypical ownerTypical frequencyEvidence artefactCommon exception
AvailabilityUptime monitoring and alertingEngineering leadContinuousStatus dashboards, alert historyAlerts fire into an unmonitored channel
AvailabilityBackup success monitoring and restore testingEngineering leadContinuous monitoring; periodic restore testsBackup logs, restore test recordsBackups green, restores never attempted
Processing integrityInput validation on critical processingEngineering leadContinuousValidation rules, rejected-input logsBad data discovered by customers first
Processing integrityReconciliation of processed outputOperations leadPer run or dailyReconciliation reports with sign-offReconciliations run, output never reviewed
Processing integrityFailed-job alerting and reprocessingEngineering leadPer eventFailure alerts and reprocessing recordsFailed jobs silently dropped
ConfidentialityData classification schemeSecurity leadReviewed annuallyClassification standard, labelled data storesScheme published, data never classified
ConfidentialityConfidentiality agreements for staff and contractorsHRPer engagementSigned agreementsContractors working without NDAs
ConfidentialityRetention and disposal schedule with executed deletionSecurity leadPer scheduleDisposal and destruction recordsRetention schedule exists, deletion never evidenced
PrivacyCurrent, accurate privacy noticeLegal or security leadReviewed on changePublished notice with version historyNotice no longer matches actual data use
PrivacyConsent records where consent is relied onEngineering leadContinuousConsent recordsConsent assumed, never recorded
PrivacyData subject request handlingLegal or security leadPer requestRequest log with response recordsRequests answered informally, nothing logged
PrivacyBreach and incident notification process, including commitments to notify vendors, data subjects and regulatorsSecurity leadPer incident; reviewed annuallyNotification procedure, incident notifications sentCommitments in contracts nobody has mapped

Common failure points

Cookie-cutter evidence. The control is copied from a template, so the evidence a company produces does not match the control as written. Auditors test the control you described, not the one you meant to describe.

Controls that only exist during the window. Evidence appears in month one and month twelve of the observation period and nowhere between. Auditors commonly expect evidence generated across the whole period. Treat any fixed staleness threshold you read online, such as a six-month limit on risk register age, as the house rule of a particular firm, not an AICPA rule (market practice note per the cluster SERP record, checked 19-08-2026).

Self-approval. The person who makes the change approves the change. This shows up in code release, access grants and vendor onboarding, and it is one of the fastest routes to an exception.

Findings that never close. Internal reviews and scans generate findings; nobody owns remediation; the same issue appears in the next examination as a repeat exception. Exception patterns reported across the market cluster on access reviews, vulnerability management and change approvals (Scrut exception triage, market practice, checked 19-08-2026).

Requirements people assume exist but do not

A fixed number of controls. No AICPA control count exists. Figures like 80 to 100 controls circulate in vendor content without any source (recorded as unsourced in the cluster SERP record, checked 19-08-2026); they describe someone’s template, not a requirement.

Individual assessment of every point of focus. The AICPA’s own text says using the criteria “does not require an assessment of whether each point of focus is addressed” (TSP section 100, par. .04, accessed 19-08-2026). Points of focus are guidance, not requirements.

Prescribed tools. No mandate exists for a particular platform, scanner, password manager or training provider. The control is the requirement; the tool is your choice.

One-to-one mapping of controls to points of focus. Not required; points of focus are not separate controls (market framing consistent with Bright Defense’s definitional treatment, checked 19-08-2026).

For the wider debunk list, including the fact that SOC 2 is an attestation rather than a certification and that the four categories beyond security are scope choices, see SOC 2 requirements.

FAQs

Is there an official SOC 2 controls list?

No. The AICPA publishes the Trust Services Criteria and points of focus, not a list of controls to implement (TSP section 100, par. .02-.04, accessed 19-08-2026). Each organisation identifies and designs controls that meet the criteria for its own system. Any numbered list you find online, including this page, is practitioner practice, not an AICPA issuance.

How many controls does SOC 2 require?

There is no fixed number. You implement enough controls to meet the criteria that fall inside your scope, and the auditor tests those controls rather than counting them. Figures such as 80 to 100 controls circulate in vendor content without sourcing; treat them as descriptions of particular templates, not rules.

Are the AICPA points of focus mandatory?

No. The AICPA describes points of focus as important characteristics of the criteria that may assist management when designing and operating controls, and states that using the criteria does not require assessing whether each point of focus is addressed (TSP section 100, par. .03-.04, accessed 19-08-2026). They are guidance you can customise or set aside where they do not fit your circumstances.

What should we hand a customer who asks for our SOC 2 controls list?

Hand them your SOC 2 report, not a list. The report is the attested artefact: it describes your system and carries an independent auditor’s opinion on your controls, which is what the customer’s security review actually needs. An internal controls list is a working document and carries no independent weight. If you are between report periods, your auditor can advise on bridging documentation.

Who should own SOC 2 controls in a software company?

A named person should own each control, typically the IT lead for access and endpoint controls, the engineering lead for change management and operations controls, HR for screening and training, and leadership for governance and risk. Shared ownership by a team is where controls quietly stop running, because nobody produces the evidence. In a small company one person may hold several controls; that is fine as long as each control has exactly one owner.

How often are SOC 2 controls tested?

In a Type 1 examination the auditor tests whether controls are suitably designed as of a specified date, often shortened to “a point in time” in SOC 2 discussions; in a Type 2 examination the auditor tests operating effectiveness across a specified period and describes the tests and results in the report (TSP section 100, par. .11, accessed 19-08-2026). The cadence of each control itself, quarterly access reviews or annual training, is your design choice. The auditor then tests whether the evidence shows you kept to it across the period.

Can we copy a SOC 2 controls list from another company?

You can use one as a starting point, but you cannot adopt it wholesale. In a SOC 2 engagement the criteria are interpreted against your commitments and system requirements, meaning the promises in your contracts and service agreements, not someone else’s (TSP section 100, par. .13-.14, accessed 19-08-2026). Copied controls usually fail on evidence: they describe a company you are not, so the records you produce never quite match.

Do we need a compliance platform to run these controls?

No platform is required by the AICPA or by the attestation standards. A platform earns its cost by automating evidence collection, tracking control owners and flagging gaps before the auditor finds them; Atoro’s partner platform is Drata, but spreadsheets and disciplined ticketing can carry a first examination. What you cannot skip is the evidence itself, however you collect it.

The honest next step

If you are building this catalogue for a first examination, the fastest working session is a control-by-control gap review against the criteria in your scope: which of these controls you already run, which run without evidence, and which are missing entirely. That review is what Atoro’s SOC 2 readiness work starts with; you can read about it on our SOC 2 service page. The guide’s usefulness is the pitch: if the tables above showed you gaps you can close yourself, close them.

Sources

  • AICPA, TSP section 100, “2017 Trust Services Criteria for Security, Availability, Processing Integrity, Confidentiality, and Privacy” (with revised points of focus 2022): par. .02-.05 (criteria structure and control design), par. .03-.04 (points of focus status, verbatim quotes), par. .11 (Type 1 vs Type 2), par. .13-.16 (commitments and system requirements), CC6.4 and CC6.8 criterion text. Accessed 19-08-2026 via full-text mirror of the AICPA publication (copyright 2017 AICPA, document 22099-349) at leaglobal.com; current edition listing confirmed on aicpa-cima.com, free account access, 19-08-2026 (per research record R26).
  • Published SOC 2 reports quoting Trust Services Criteria text verbatim: metr.org 2025 SOC 2 Type 1 report; profisee.com 2024 Type 2 report. Accessed 19-08-2026 (per research record R26).
  • Market practice (tier 2, named and dated, used for practice observations only, never as AICPA rules): soc2auditors.org SOC 2 controls list page (evidence and gap column format); scrut.io SOC 2 common criteria page (exception-risk triage); brightdefense.com SOC 2 controls list page (criteria vs controls vs points of focus definitions). All checked 19-08-2026 (per research record R27).
  • Atoro supplied facts: positioning as an Irish AI governance and cyber compliance consultancy for software companies; first consultancy in Europe certified against ISO/IEC 42001, certificate AIMS-AT-120224, A-LIGN under ANAB; “more than 200 certifications delivered” (atoro.io, checked 08-08-2026).