
Compliance
SOC 1 vs SOC 2: Which Report Your Customer's Auditor Needs
SOC 1 vs SOC 2 turns on who reads it: SOC 1 serves your customer's financial-statement auditor, SOC 2 serves security reviewers. Here is what pulls you in.
Most comparisons open with a table: SOC 1 is financial, SOC 2 is security. Not wrong, but it answers the wrong question. The useful one is who will read the report, and what they will do with it.
This piece works from the texts that reader applies and maps them onto systems a platform team runs. If you know the certificate-versus-attestation distinction from ISO 27001 vs SOC 2, this is the other axis: two attestation reports, two different audiences.
The question is whose auditor reads it
The AICPA defines SOC 1 by its reader. It is "an examination of controls at a service organization that are likely to be relevant to user entities' internal control over financial reporting," and the reports are "specifically intended to meet the needs of entities that use service organizations (user entities) and the CPAs that audit the user entities' financial statements (user auditors)." Even the AICPA guide's title names that reader: "Reporting on an Examination of Controls at a Service Organization Relevant to User Entities' Internal Control Over Financial Reporting."
The SOC 2 publication is named by subject instead: "Reporting on an Examination of Controls at a Service Organization Relevant to Security, Availability, Processing Integrity, Confidentiality, or Privacy." The AICPA's SOC 2 page — as archived in January 2023, since the current page redirects to a landing page — described those reports as meeting "the needs of a broad range of users," and listed oversight of the organization, vendor management programs, internal governance and regulatory oversight among their uses.
So the one-line test: a customer's financial-statement auditor wants a SOC 1. A security questionnaire or vendor-risk team wants a SOC 2. A prospect reading your website gets a SOC 3, which the AICPA describes as covering the same categories as SOC 2 without "the same level of detail" and therefore general use, able to be "freely distributed."
None of the three is a certification. Each is a CPA's examination report — an attestation engagement — with an opinion dated to a point or a period, and a reader who decides what it is worth.
What each report actually attests to
The two reports sit under different attestation standards. The AICPA's Technical Q&A section 9561, published in September 2026, states the split plainly: a SOC 1 examination is "performed in accordance with AT-C section 320, Reporting on an Examination of Controls at a Service Organization Relevant to User Entities' Internal Control Over Financial Reporting," and a SOC 2 examination "in accordance with AT-C section 205, Assertion-Based Examination Engagements." That AT-C 205 title is itself recent: SSAE No. 21, issued in September 2020, superseded the SSAE 18 version of AT-C 205, then titled Examination Engagements.
The criteria differ in kind. In a SOC 1 the service organization's description states the control objectives, and the service auditor opines against them — the IAASB's overview of ISAE 3402, the international counterpart, describes a type 1 opinion on whether "the controls related to the control objectives stated in the description were suitably designed." You designate them, though AS 2601.35 lets outside parties such as regulators or a user group set them.
A SOC 2 is measured against fixed criteria. The 2017 Trust Services Criteria (cited here from an archived AICPA copy; the current edition keeps these criteria with revised points of focus, 2022) consist of criteria common to all five categories plus additional criteria for availability, processing integrity, confidentiality and privacy, and the common criteria alone are "suitable for evaluating the effectiveness of controls to achieve an entity's system objectives related to security." Security is the floor.
- Type 1 — the description and the suitability of design, as of a date. The archived TSC states a type 1 SOC 2 report contains neither "an opinion on the operating effectiveness of controls nor a detailed description of tests of controls performed by the service auditor and the results of those tests."
- Type 2 — design plus operating effectiveness over a period, with the auditor's tests and results. ISAE 3402 frames it the same way: the auditor "also opines on whether those controls operated effectively throughout a specified period."
- Distribution — the archived AICPA SOC 2 page said "Use of these reports are restricted." SOC 3 is the general-use variant.
When your platform is in SOC 1 territory
The cleanest test sits on the user auditor's side. PCAOB AS 2601.03 — the issuer standard, from the PCAOB compilation for fiscal years beginning on or after 15 December 2024 — says a service organization's services are part of a company's information system if they affect "any of the following" — five effects. Two are the obvious ones: "the procedures, both automated and manual, by which the entity's transactions are initiated, recorded, processed, and reported from their occurrence to their inclusion in the financial statements," and the related accounting records. The other three reach further: significant classes of transactions, other events and conditions significant to the financial statements, and the financial reporting process itself, including significant accounting estimates and disclosures. A consolidation, estimation or disclosure system can be in SOC 1 territory without recording a single transaction. Its examples include "application service providers that provide packaged software applications and a technology environment that enables customers to process financial and operational transactions."
It also carves an exclusion: the provisions do not apply where services are "limited to executing client organization transactions that are specifically authorized by the client, such as the processing of checking account transactions by a bank or the execution of securities transactions by a broker." Private-company audits follow the AICPA's AU-C 402, not covered here.
Applied to a platform team's systems — generic patterns from the standard's text, not a ruling on any estate:
- Clearly in: an invoicing engine that issues invoices in the customer's name; a ledger service that is the customer's book of record; payroll calculation; a reconciliation pipeline whose output a customer posts to its general ledger. Your own fee invoices to customers are not in scope merely because customers pay them.
- Clearly out of SOC 1: a support ticketing system, a documentation site, product analytics. These are SOC 2 territory if anything.
- The exclusion case: a payout service that only executes transfers each customer explicitly instructs looks like the broker example. Record why; the user auditor may disagree.
- The borderline case: a managed platform on which the customer runs its own ledger. You initiate and record nothing, but your access and operations controls sit beneath an application that does. Whether that pulls you in is the user auditor's judgement, which is why the question reaches infrastructure providers at all.
Infrastructure under an in-scope application is not out just because it holds no invoices. AS 2110 Appendix B .B1 has the auditor understand "the IT general controls that are important to the effective operations of the automated controls" — so the cluster, CI system and database under your invoicing engine carry the general controls the report will lean on. Keep the scoping decision as data, versioned beside the code:
# report-scope.yaml
# One entry per system. soc1_test records the five AS 2601.03 effects:
# the first boolean covers two of them (transaction procedures and the
# related accounting records); the second covers the other three
# (significant classes of transactions, other events and conditions
# significant to the financial statements, and the financial reporting
# process incl. estimates and disclosures). Any true means in scope
# unless the client-authorized-execution exclusion applies.
# Control objectives are yours to define, so they are named here.
period:
start: "2026-01-01"
end: "2026-12-31"
systems:
- name: customer-invoicing
namespace: invoicing
soc1_test:
initiates_records_processes_or_reports_customer_transactions: true
affects_classes_events_or_financial_reporting_process: true
limited_to_client_authorized_execution: false
reports: [soc1, soc2]
soc1_control_objectives:
- "CO-03: invoices issued in the customer's name are complete and accurate"
- "CO-07: changes to rating logic are authorized, tested and approved"
soc2_categories: [security, processing-integrity]
- name: ledger
namespace: ledger
soc1_test:
initiates_records_processes_or_reports_customer_transactions: true
affects_classes_events_or_financial_reporting_process: true
limited_to_client_authorized_execution: false
reports: [soc1, soc2]
soc1_control_objectives:
- "CO-04: postings are recorded completely, once, in the right period"
- "CO-09: write access to ledger data is restricted to authorized identities"
soc2_categories: [security, availability, processing-integrity]
- name: payout-executor
namespace: payouts
soc1_test:
initiates_records_processes_or_reports_customer_transactions: true
affects_classes_events_or_financial_reporting_process: true
limited_to_client_authorized_execution: true
reports: [soc2]
rationale: "Executes only transfers each customer instructs; confirm with user auditors."
soc2_categories: [security, availability]
- name: platform-cluster
namespace: "*"
soc1_test:
initiates_records_processes_or_reports_customer_transactions: false
affects_classes_events_or_financial_reporting_process: false
limited_to_client_authorized_execution: false
reports: [soc1-supporting, soc2]
rationale: "Hosts customer-invoicing and ledger; its change, access and operations controls are ITGCs for them."
soc2_categories: [security, availability]How the user auditor consumes a SOC 1
AS 2601 uses older vocabulary: a "report on controls placed in operation" is what is now called type 1, and one that adds "tests of operating effectiveness" is type 2. AS 2601.24 settles who picks: "The type of engagement to be performed and the related report to be prepared should be established by the service organization" — though it calls discussion with user organizations "advisable." The decision is yours.
Type 1 is where this goes wrong. AS 2601.24a says a type 1 report may help the user auditor plan, but is "not intended to provide the user auditor with a basis for reducing his or her assessments of control risk below the maximum." In the integrated audit of internal control, AS 2201 Appendix B is blunter: a report without tests of controls and an operating-effectiveness opinion "does not provide evidence of operating effectiveness."
With a type 2 in hand, AS 2201.B21 has the user auditor weigh three things: the period covered "and its relation to the as-of date of management's assessment"; the scope, applications and controls tested; and the results and opinion. Two consequences follow for the provider:
- Complementary controls flow back to your customer. Under .B22, the duty starts only when the service auditor's report contains a qualification — the standard's term for this caveat, not a modified opinion — that the stated control objectives might be achieved only if the customer applies controls contemplated in your system's design; then "the auditor should evaluate whether the company is applying the necessary procedures." Each such complementary user entity control is a sentence in your report that becomes work in your customer's audit. Write few, precisely.
- The gap between your period and their year-end is theirs to bridge. Under .B24, when "a significant period of time has elapsed" between your test period and management's assessment date, the user auditor performs additional procedures, including asking about changes such as "errors identified in the service organization's processing." Align your period end with your customers' fiscal years where you can.
Having no report is not an exemption. AS 2201.B20 lists the user auditor's alternatives: obtain a service auditor's report with tests of operating effectiveness, or an agreed-upon-procedures report describing tests of controls; test the customer's own controls over your activity, such as "independent re-performance of selected items" or reconciliation of your output to source documents; or "performing tests of controls at the service organization." That last route means each customer's auditors testing your controls themselves.
The ITGC evidence both reports share
AS 2201.B29 names the general controls a user auditor cares about when relying on an automated control year to year: "general controls over program changes, access to programs, and computer operations." As engineering judgement, not quotation, those same three activities are the backbone of any security examination, SOC 2 included. Build one evidence source per control and point both reports at it — the same discipline as SOC 2 on self-hosted Kubernetes, extended to a second reader.
Program change: the population, not the sample
Auditors sample from a population, so a population with holes is a finding before any sample is drawn. The common mistake is git log --merges, which the git documentation describes as "exactly the same as --min-parents=2": it drops every single-parent commit — squash merges, rebase-and-merge commits, fast-forwards and most direct pushes. Walk the release branch with --first-parent instead and flag anything without a pull-request number. --since stops at the first older commit, so non-monotonic committer dates silently drop in-period commits; --since-as-filter (Git 2.37 or later) walks the whole range:
#!/usr/bin/env bash
# change-population.sh: every commit that landed on a release branch during an
# audit period, however it arrived (merge, squash, rebase, direct push).
# Usage: change-population.sh <repo-path> <branch> <period-start> <period-end>
# dates as YYYY-MM-DD, interpreted in UTC, end date inclusive.
set -euo pipefail
if [[ $# -ne 4 ]]; then
echo "usage: $0 <repo-path> <branch> <period-start> <period-end>" >&2
exit 64
fi
repo=$1
branch=$2
start=$3
end=$4
# RFC 4180 quoting: wrap in double quotes, double any embedded quote.
csv() {
local v=${1//\"/\"\"}
printf '"%s"' "$v"
}
printf 'sha,commit_date,author,committer,subject,parent_count,pr_ref,flag\n'
# --first-parent: walk only the branch's own history, one row per change that
# landed. Deliberately NOT --merges, which keeps only commits with 2+ parents
# and silently drops squash merges, rebases, fast-forwards and direct pushes.
git -C "$repo" log --first-parent "$branch" \
--since-as-filter="${start} 00:00:00 +0000" --until="${end} 23:59:59 +0000" \
--date=iso-strict \
--format='%H%x1f%cd%x1f%an <%ae>%x1f%cn <%ce>%x1f%s%x1f%P' |
while IFS=$'\x1f' read -r sha cdate author committer subject parents; do
read -r -a parent_list <<<"$parents"
pr_ref=""
if [[ $subject =~ ^Merge\ pull\ request\ \#([0-9]+) ]]; then
pr_ref="#${BASH_REMATCH[1]}"
elif [[ $subject =~ \(\#([0-9]+)\)$ ]]; then
pr_ref="#${BASH_REMATCH[1]}"
fi
flag=""
if [[ -z $pr_ref ]]; then
flag="NO_PR_REFERENCE"
fi
printf '%s,%s,%s,%s,%s,%s,%s,%s\n' "$sha" "$cdate" \
"$(csv "$author")" "$(csv "$committer")" "$(csv "$subject")" \
"${#parent_list[@]}" "$pr_ref" "$flag"
doneThe PR reference comes from GitHub's default merge and squash subject lines, so rebase-merged commits flag too; resolve those through your forge's API before calling them exceptions. Two limits belong in the control narrative: the date filter uses committer dates, and git shows what landed on the branch, not what ran in production. Reconcile the population against deploy records before handing it over.
Access to programs: grants are not exclusivity
apiVersion: v1
kind: ServiceAccount
metadata:
name: ledger-deployer
namespace: ledger
---
apiVersion: rbac.authorization.k8s.io/v1
kind: Role
metadata:
name: ledger-deploy
namespace: ledger
rules:
- apiGroups: ["apps"]
resources: ["deployments", "statefulsets"]
verbs: ["get", "list", "watch", "create", "update", "patch", "delete"]
- apiGroups: ["batch"]
resources: ["jobs"]
verbs: ["get", "list", "watch", "create", "update", "patch", "delete"]
- apiGroups: [""]
resources: ["configmaps", "services"]
verbs: ["get", "list", "watch", "create", "update", "patch", "delete"]
---
apiVersion: rbac.authorization.k8s.io/v1
kind: RoleBinding
metadata:
name: ledger-deploy
namespace: ledger
subjects:
- kind: ServiceAccount
name: ledger-deployer
namespace: ledger
roleRef:
apiGroup: rbac.authorization.k8s.io
kind: Role
name: ledger-deployThis manifest is design, not evidence. The Kubernetes documentation is explicit that RBAC "permissions are purely additive (there are no 'deny' rules)," and that a RoleBinding grants only within its namespace while a ClusterRoleBinding grants cluster-wide. Nothing here removes write access from a cluster-admin binding, a RoleBinding to the built-in edit role, or a CI identity bound elsewhere.
"Only authorized identities can write to the ledger" is therefore a separate claim, evidenced each period: list every RoleBinding in ledger and every ClusterRoleBinding, then resolve each subject with kubectl auth can-i --list -n ledger --as=<subject> — ServiceAccounts as system:serviceaccount:<namespace>:<name>, Group subjects via --as-group, which kubectl accepts only alongside --as — so pair it with an identity that holds no grants of its own, or the result overstates what the group alone can do. The dated inventory is the evidence; the manifest is the design. The API server audit trail described in Kubernetes audit logs into a SIEM you operate shows who called the API. Computer operations is a separate population: job outcomes, backup and restore records and incident tickets come from the scheduler, backup tool and ticketing system that produce them.
Why some providers need both, and the order to get them
Who is asking decides the order. Section 404(a) of Sarbanes-Oxley requires an issuer's annual report to contain management's assessment of internal control over financial reporting; 404(b) requires the issuer's audit firm to attest to that assessment, except for emerging growth companies; and 404(c) exempts issuers that are neither large accelerated nor accelerated filers from 404(b). A customer whose auditor attests under 404(b) is the reader with the clearest need for a type 2 SOC 1.
The tempting shortcut is a SOC 2 with processing integrity. The archived 2017 TSC defines processing integrity as system processing that is "complete, valid, accurate, timely, and authorized to meet the entity's objectives" — the provider's objectives. A SOC 1's control objectives are framed around your customers' financial reporting. AS 2201.B20 leaves the evidence route to the user auditor's judgement, so a SOC 2 may sometimes be accepted; SOC 1 is the report designed for that reader.
Failure modes, exit ramps and the long game
- Type 1 to a 404(b) customer. Under AS 2201 Appendix B it provides no evidence of operating effectiveness. The user auditor then obtains a report with tests of operating effectiveness or agreed-upon procedures, tests the customer's own controls over your service, or tests controls at your organization. Only the request for a type 2 and the on-site tests involve you; the first visible sign is often a request to send auditors in.
- Scope drift between periods. A system added to the ledger path mid-year and absent from the description is a gap the user auditor finds under .B21. Diff
report-scope.yamlat each period end. - Unnamed dependencies. If the ledger depends on another provider whose controls your report leaves out, that is a hole your customers' auditors must fill some other way — the .B20 alternatives apply to it just as they apply to you. Name it in the description and in the evidence plan.
- Choosing an auditor without checking credentials. The AICPA has publicly said it is looking into anonymously published allegations about a compliance vendor offering SOC services, and will act if auditors were not performing audits to professional standards, not enrolled in peer review, or unlicensed. Check licensing and peer review before signing.
The exit ramp is evidence you own. If the change population, binding inventory and operations records are generated from your repositories and your cluster, they survive a change of audit firm, a move between SOC 1 and ISAE 3402 for customers outside the United States, and a criteria revision. The AICPA published exposure drafts of proposed 2027 SOC 2 Trust Services Criteria and Description Criteria on 26 October 2026; they are proposals, not effective criteria, and evidence that regenerates from source absorbs a remapping more cheaply than hand-assembled evidence.
The same logic runs through PCI DSS audit scope and third-party risk when one vendor runs your platform: scope is an argument you can re-derive, and assurance is a by-product of controls you already run.
§FAQ/Common questions
Frequently asked
What is the difference between a SOC 1 and a SOC 2?
The reader. A SOC 1 is an examination of controls at a service organization likely to be relevant to user entities' internal control over financial reporting, performed under AT-C section 320 and intended for those customers and the CPAs who audit their financial statements. A SOC 2 is performed under AT-C section 205 against the AICPA's Trust Services Criteria — security, availability, processing integrity, confidentiality and privacy — and serves a broad range of users such as vendor-management and oversight reviewers. In a SOC 1 the service organization states its own control objectives; in a SOC 2 the criteria are fixed.
Does a cloud or platform provider need a SOC 1?
Only if its services are part of a customer's information system for financial reporting. PCAOB AS 2601.03, the issuer standard, includes services that affect how a customer's transactions are initiated, recorded, processed and reported, and names application service providers that let customers process financial transactions. It excludes services limited to executing transactions the customer specifically authorized. A platform running a customer's ledger, payroll or invoicing in the customer's name is usually in; your own fee invoices to that customer are not; infrastructure under a customer's own financial application is a judgement call the customer's auditor makes.
Is a SOC 1 type 1 report enough for a SOX audit?
Not as evidence of operating effectiveness. AS 2601.24a says a report on controls placed in operation (type 1) is not intended to give the user auditor a basis for reducing assessed control risk below the maximum, and AS 2201 Appendix B states such a report does not provide evidence of operating effectiveness. Under AS 2201.B20 the auditor then needs a type 2, an agreed-upon-procedures report describing relevant tests of controls, tests of the customer's own controls over the service, or tests of controls at the service organization.
Can a SOC 2 with processing integrity replace a SOC 1?
Sometimes, at the user auditor's discretion, but it is not designed to. The 2017 Trust Services Criteria define processing integrity against the service organization's own objectives, while a SOC 1's control objectives address customers' internal control over financial reporting. AS 2201.B20 leaves the choice of evidence to the user auditor, so ask the customer's auditor rather than assuming.
Is SOC 1 or SOC 2 a certification?
Neither. Both are CPA examination reports under AICPA attestation standards — AT-C 320 for SOC 1 and AT-C 205 for SOC 2 — carrying an auditor's opinion as of a date (type 1) or over a period (type 2). The archived AICPA SOC 2 page described their use as restricted; the general-use, freely distributable variant is SOC 3.
Executive Briefing
Thirty minutes to clarify your infrastructure risk
Walk us through your vendor footprint and regulatory constraints. We will tell you honestly where sovereignty creates leverage — and where it does not. No pitch deck. No obligation.