Responsibility matrix
This page is for a company preparing for a SOC 2 audit, or meeting its duties under South Africa's Protection of Personal Information Act (POPIA). It says, control by control, what Hosting does for the part of your system we run, what you must do yourself, and where we have not closed a gap yet.
Three things first, stated plainly:
- Hosting holds no SOC 2 report today (2026-11-11). Your auditor will treat us as a subservice organization and either ask for our report, which does not exist, or carve us out of your report.
- Hosting with us does not make you compliant with SOC 2 or POPIA. Only your auditor issues your SOC 2 report, and only you can meet your POPIA duties.
- Under POPIA you are the responsible party and Hosting is your operator. You decide why and how personal information is processed and you answer to the Information Regulator and to the people whose information you hold. We process it on your behalf. Where your data is held and who our subprocessors are is on Data location and subprocessors.
The matrix
Owner is who operates the control: Hosting, you (the client), or shared — both of us do a part. The criteria are SOC 2's 2017 Trust Services Criteria (TSC); the sections are POPIA's.
| Area | Owner | TSC | POPIA | What Hosting does | What you must do |
|---|---|---|---|---|---|
| Logical access to Hosting | Shared | CC6.1–CC6.3 | s19 | Company membership and roles decide who can use the panel and what each role may do; every action is checked inside your company. A Company Admin can require two-factor sign-in or a passkey for every member. Our staff reach your company only by signing in on your behalf, with the operator named on every change (Your company, two-factor sign-in) | Invite and remove your own members, choose their roles, turn on the two-factor requirement, and review who has access each quarter |
| Access to environments | Shared | CC6.1, CC6.6 | s19 | SFTP and SSH logins are separate per-environment users you create, with keys, passwords and a read-only mode (SFTP and SSH access) | Create and remove logins yourself. Removing a member does not remove their SFTP/SSH login |
| Your application's own users | You | CC6.1–CC6.3 | s19 | Nothing: we do not administer the users inside your application | Manage your application's accounts, administrators and their second factor |
| Audit trail | Shared | CC4.1, CC7.2 | s19 | Every change inside Hosting for your company is recorded with who made it and when, including the operator behind a change made on your behalf. The trail is kept for 365 days; closure, cancellation and deletion records are archived for longer (Activity) | Read the trail when reviewing access or a change. What happens inside your application is your own log |
| Encryption in transit | Shared | CC6.1, CC6.7 | s19 | Every published address is served over HTTPS with a certificate we request (Web addresses) | Use TLS for your application's own outbound connections; a database connection is encrypted only when your application asks for it (Configuration) |
| Encryption at rest | Shared | CC6.1, C1.1 | s19 | File and database volumes are provisioned from an encrypted storage class, and file backups are sealed in an encrypted restic repository (Encryption at rest, Backups). Gap: database backups (PostgreSQL base backups and WAL, MySQL/MariaDB dumps) sit in object storage with no provider-side encryption configured, and object storage you buy has no default encryption (Object storage) | Encrypt anything in object storage that needs it, and anything you copy out of the platform |
| Physical and facility security | Hosting, through Hetzner | CC6.4, A1.2 | s19 | Our servers and object storage are at Hetzner in Falkenstein, Germany. Hetzner holds ISO/IEC 27001:2022 and a BSI C5 Type 2 attestation (Data location and subprocessors) | Decide whether your auditor accepts Hetzner's certificates for a carved-out subservice organization |
| Backups and recovery | Shared | A1.2, A1.3 | s19 | Two backup tracks, files and databases, keeping one restore point per day: 7 days included, up to 30 by agreement. A restore is asked for through support (Backups and restores, Restore patterns). Gap: no numeric recovery time or data-loss bound is committed | Choose your retention, keep an independent copy of critical data, and test a restore of your own application |
| Change management | Shared | CC8.1 | s19 | Our platform and the runtime images it pins change only as versioned releases; recovery is a new release, never a downgrade. Your application deploys from an exact commit into a retained release (Applications) | Review, test and approve changes to your own code before you deploy them |
| Availability incidents | Shared | CC7.3–CC7.5 | — | When an application cannot use its database, an incident is confirmed after sustained failure and your company is notified once on confirmation and once on recovery. Gap: no availability commitment (SLA) is offered | Keep your contact details current, and run your own incident response for your application |
| Security incidents | Shared | CC7.3–CC7.5 | s21(2), s22 | When we have reasonable grounds to believe your data was accessed by an unauthorized person, your owners are told at once in the panel and by email, with what we know and what we did. The notice stays on the Compliance page | Plan who in your company notifies the Regulator and the people affected |
| Vulnerabilities | Shared | CC7.1 | s19 | We patch the platform and the runtime images we pin. Your container images are scanned in the background and the results shown to you; a finding never blocks a deploy (Container images) | Patch your own code and its dependencies, and act on your image's scan results |
| Monitoring | Shared | CC7.2 | s19 | We monitor the platform's own services, and the database-connectivity incidents above. Your company can set its own disk free-space alert rules | Read your application's logs (Logs), set your alert rules and monitor your application itself |
| Offboarding and deletion | Shared | C1.2 | s14 | Cancellation shows what stops, the service-end date and the recoverable-until date before you confirm; the environment is then deleted, and its backups are kept until the recoverable-until date or destroyed at once if you choose (Cancelling hosting) | Remove members who leave, take what you need out before the recoverable-until date, and delete your own backup bucket yourself — deleting an environment does not reach into it |
| Information Officer | You | — | s55–s56 | Nothing: this is your registration | Register your Information Officer with the Regulator |
| Privacy notice | You | P1.1 | s18 | Where your data is held, for you to cite (Data location and subprocessors) | Publish a notice to your users that says where their data is held |
| Data-subject requests | You | P5.1, P5.2 | s23–s25 | Your company can take a copy of a database and reach its files over SFTP | Find, correct and delete a person's information in your own application |
| Retention of your data | You | C1.2 | s14 | Retention and deletion dates as above | Decide how long you keep personal information inside your application |
| Cross-border transfer | You | — | s72 | The facts and the basis for you to adopt (Data location and subprocessors) | Record the transfer and its basis in your own POPIA documentation |
| Operator contract | Shared | — | s20–s21 | Our operator terms are offered on the Compliance page and record who accepted which version, and when | Have your own legal adviser review the terms, accept them, and keep the generated copy on file |
SOC 2: what your auditor will ask of you
A SOC 2 report on your system lists complementary user entity controls (CUECs): controls you, the user of our service, must operate for our part to work. From the matrix, yours are:
- Review your company's members and their roles at least quarterly, and remove people who have left.
- Turn on the company's two-factor requirement, and make sure every member has enrolled.
- Create, review and remove SFTP/SSH logins for each environment.
- Administer your application's own users, its administrators and their second factor.
- Own your application code: review changes before deploying, and keep it and its dependencies patched.
- Act on your container image scan results.
- Choose your backup retention, keep an independent copy of critical data, and test your own restore.
- Encrypt data you put in object storage or take out of the platform when it needs encryption.
- Run incident response for your application, and keep your notification contacts current.
- Delete data you hold elsewhere, including your own backup bucket, when you offboard.
Hosting is a subservice organization in your report, and Hetzner is ours. Because we hold no SOC 2 report, your auditor will most likely use the carve-out method: our controls are excluded from the report and your auditor monitors us another way, for example from this page. No authoritative source says an auditor accepts Hetzner's C5 or ISO 27001 in place of a SOC 2 report; that is your auditor's call.
POPIA: the duties only you can do
As the responsible party, these are yours. For each, what Hosting provides to help.
| Your duty | The Regulator's material | What Hosting provides |
|---|---|---|
| Register an Information Officer (s55–s56). It is compulsory and free, and reporting a compromise needs it first | eServices portal, Information Officer guidance note | Nothing — the registration is yours alone |
| Tell your users what you collect, why, and where it is held (s18) | POPIA, Act 4 of 2013 | Where your data is held, on Data location and subprocessors |
| Answer access and correction requests (s23–s25) | POPIA Regulations and their 2025 amendments (the request forms) | A copy of a database on request (Taking a copy of a database) and SFTP access to your files (SFTP and SSH access); finding the person's records in your application is yours |
| Decide how long you keep it (s14) | POPIA, Act 4 of 2013 | Backup retention you choose, and the dates below |
| Notify a security compromise to the Regulator and the people affected (s22). Every compromise is reported, through the portal | Reporting through the portal, step-by-step guide, fact sheet | We tell your owners immediately, in the panel and by email, with what you need for your own report |
| Justify the cross-border transfer (s72) | POPIA, Act 4 of 2013; no cross-border guidance or adequacy list had been published on 2026-11-10 | The facts and our reasoning for you to adopt, on Data location and subprocessors |
| Keep a PAIA manual (PAIA s51) | PAIA manual template for a private body | Nothing — the manual is yours |
| Get prior authorisation before sending special or children's information to a country without adequate protection (s57(1)(d)) | Prior authorisation form | The facts on where it is held. Take your own advice if you hold such information |
What happens to deleted data in backups
This decides what you may tell a person who asked for their information to be erased. When an environment is cancelled, it is deleted at the service-end date you chose. Its backups — files and databases — are kept until the recoverable-until date: the service end plus the retention days in force, which is 7 days unless you agreed more, up to 30. They are then pruned and the data is gone. If you choose Delete everything now, the retained backups are destroyed at once instead. A copy in your own backup bucket is yours and is never deleted by us. See Cancelling hosting.
Deleting one record inside your application does not reach into earlier restore points: it leaves the backups when they age out of your retention.
Next
- Where your data is held and who our subprocessors are: Data location and subprocessors.
- Questions about this page: Support.