Trust center
Trust
Where does superdiffs stand today?
superdiffs is in public beta. That has consequences a security reviewer needs stated directly rather than inferred: there is no SOC 2 report and no audit under way, no completed penetration test, no SSO or audit log export, and several subprocessors listed below are ones we intend to use rather than ones we use today. Each is marked.
What is already true is architectural, because it was designed in rather than certified after: your code is analyzed in a disposable sandbox, the credential that reaches it is a one-hour read-only token for a single repository, and nothing you give us trains a model.
Does superdiffs train on my code?
No. We do not train models. We do not fine-tune on customer code, and we do not send customer code to a provider under terms that permit training on it.
Which model provider sees my code?
Anthropic: Model provider (Claude) for review, verification and judge passes. Region: United States.
Zero data retention at the provider: Not in force. No zero-retention amendment with our model provider has been executed. Until one is, provider-side retention is the provider’s published default, and this page will name the agreement when it exists rather than imply one we do not have. What reaches the provider, and what is redacted before it does, is on the subprocessor list and in the six facts below.
How is my code isolated?
In six facts:
Analysis runs in a disposable sandbox
Your code is checked out inside a per-review microVM that is destroyed when the review ends. Tokens that can post, platform secrets and model keys never enter it. Outbound network access from the sandbox is not yet restricted to an allow-list; that is the next hardening step, and this line will change when it ships.
One read-only, repository-scoped token
The only credential that crosses into the sandbox is an installation token limited to a single repository, contents-read, valid for one hour. It cannot post, cannot read a second repository, and cannot be widened by the code that requests it — the permission set is a frozen literal, and a test asserts that nothing reachable from that module can mint a posting token.
The signing key stays with one service
Today the GitHub App private key is a secret on our hosting provider, read only by the api’s token broker and never passed into a sandbox. A signer backed by a cloud key-management service is implemented and is the planned hardening step, so that a compromised process would not hold the key itself.
Instructions come from people, not from text
Bot commands are honored only from a comment written by someone with write access, checked against the permission API at the moment of use. A command in a code comment or in the diff itself is inert no matter how it is phrased, and a pull-request description can carry exactly one — ignore — which can only stop a review.
Both directions are sanitized
Untrusted repository content is stripped of invisible and tag-smuggled characters and structurally delimited before a model sees it, and detected credentials are redacted from every prompt before it is sent to a model provider. Everything we post is stripped of image channels, raw HTML, and foreign links, so a review comment cannot become an exfiltration channel.
No training on customer code
Customer code is used to produce your review and nothing else. It is not used to train or fine-tune any model, and cross-customer quality work uses metadata about our own output — which rule fired, whether it was addressed — never code.
The engineering detail behind each of these — permission scopes, the sanitization rules, the threat model — is in the security documentation.
Does superdiffs have SOC 2?
Not yet. No audit has started, the compliance tooling is not in place, and no report exists today. Type I is planned after beta and targeted before general availability, with the Type II observation window after it. We will not describe the program as "SOC 2 compliant" until an auditor has issued a report. The rest of the compliance status:
| Item | Status | Detail |
|---|---|---|
| SOC 2 | Not started | No audit has started, the compliance tooling is not in place, and no report exists today. Type I is planned after beta and targeted before general availability, with the Type II observation window after it. We will not describe the program as "SOC 2 compliant" until an auditor has issued a report. |
| Zero data retention (model providers) | In progress | Not in force. No zero-retention amendment with our model provider has been executed. Until one is, provider-side retention is the provider’s published default, and this page will name the agreement when it exists rather than imply one we do not have. |
| Zero-retention mode (our systems) | In progress | Not available. An enterprise-tier flag meant to suppress evidence-snippet storage and learnings, so nothing derived from your code outlives the review. The entitlement exists in code; the enforcement path is not finished, and nothing is sold as zero-retention until it is. |
| GDPR | In progress | The Data Processing Addendum is published and available for signature, with standard contractual clauses for transfers outside the EEA. Subject-access and erasure requests go to privacy@superdiffs.com and are answered by a person. |
| Penetration test | Not started | No external test has been done. One covering the sandbox boundary and the token broker is planned before general availability; the summary letter will be linked here when it exists. |
| Training on customer code | N/A | We do not train models. We do not fine-tune on customer code, and we do not send customer code to a provider under terms that permit training on it. |
Data handling and retention
| Data | Where it lives | Retention | Under zero-retention mode (not built) |
|---|---|---|---|
| Repository checkout (working copy) | Sandbox instance | Fetched straight into the sandbox for the review and destroyed with it. Nothing is staged in object storage. | Same |
| Finding evidence snippets and suggested patches | Postgres | Life of the finding — they are what lets us re-anchor a comment after a force-push without re-reading the old commit. Deleted on uninstall. | Not stored; findings degrade to file-level anchors after a force-push |
| Review and pull-request metadata | Postgres | Kept for usage history and billing disputes, for at most 13 months: an hourly job deletes reviews and their findings once they pass that age. | Same — no code content is involved |
| Learnings (embedded team preferences) | Postgres (vector index) | Until deleted by an admin in the dashboard, or on uninstall | Disabled |
| Prompts and completions | Model provider | The provider’s default — no zero-retention amendment has been executed | Zero retention once an amendment is in force |
| Code-derived data, on uninstall | Postgres | Evidence snippets, suggested patches, learnings and cached configuration are deleted when the app is uninstalled, in one transaction that writes an erasure record. The rest of the installation’s data (repositories, pull requests, reviews, findings) is erased 30 days after uninstall unless you reinstall, and each batch writes an erasure record. Billing and audit records are kept. | Same |
Zero retention, stated precisely
“ZDR” is used loosely in this market to mean two different things, and the difference matters:
- At the model provider — whether prompts are retained by the provider after the response. This requires an executed amendment with the provider. None has been executed, so it is not in force today: treat provider-side retention as the provider’s published default.
- In our systems — whether we store anything derived from your code after the review. This is a planned enterprise-tier mode that suppresses evidence-snippet storage and learnings. The entitlement exists in code; the enforcement path is not finished.
Neither is a checkbox we have ticked, and a vendor at our stage telling you it has both is worth a follow-up question.
Deletion
Uninstalling the GitHub App deletes, in one transaction, the evidence snippets and suggested patches stored with your findings, your learnings and the cached configuration, and writes an erasure record with the counts. The rest — repositories, pull requests, reviews and findings — is erased 30 days after uninstall unless you reinstall, and review metadata is never kept longer than 13 months; an hourly job enforces both and records each deletion. To have anything deleted sooner, or for any request under GDPR, write to privacy@superdiffs.com.
Who are the subprocessors?
Everything that could touch customer data, whether or not it does today. The status column is not decoration: a list that quietly mixes vendors we use with vendors we plan to use is the kind of small dishonesty that makes the rest of a trust page worthless.
| Subprocessor | Status | Purpose | Region |
|---|---|---|---|
| Anthropic | In use | Model provider (Claude) for review, verification and judge passes. | United States |
| Amazon Web Services | In use | Underlying infrastructure for the E2B sandbox fleet. Our own application hosting is Railway, listed separately. | United States, Europe and Japan (sandbox regions) |
| Hetzner | Planned | Contested alternative for the sandbox fleet. Not in use. | Germany / Finland |
| E2B | In use | Sandbox provider: the disposable machines that check out code and let verification read and search it. | United States (region selectable per review) |
| Temporal Cloud | Planned | Under consideration for durable review orchestration. Not in use; the review queue is Redis-backed today. | European Union |
| Paddle | In use | Merchant of record: payments, subscriptions, invoicing and sales tax. | United Kingdom / United States |
| PostHog | Planned | Product analytics for the web application. | European Union |
| Sentry | Planned | Error and performance monitoring. The integration is in the code but no DSN is configured, so nothing is transmitted today. | European Union |
| Railway | In use | Application hosting for the api and web services, and the managed Postgres and Redis instances. | United States and Europe (region selectable) |
| Resend | In use | Transactional email: the welcome message, plan and usage notices, and trial reminders. | United States |
| Vanta | Planned | Compliance automation and continuous control monitoring. | United States |
What reaches each vendor is on the subprocessor list. Material changes to it are announced 30 days before they take effect, to the billing contact and in the changelog. Customers on a signed DPA may object within that window.
Egress addresses
superdiffs has no static egress IP addresses yet, so there is nothing stable to put in a firewall rule. When there are, the ranges will be published here and in a machine-readable file at /.well-known/egress.json, with 30 days notice before any change.
Machine-readable location: /.well-known/egress.json. It lists no ranges, because we would rather publish nothing than publish an address that turns out to be wrong in your firewall.
If your GitHub is behind an IP allow-list, tell us — that is what decides whether we move to static addresses at all.
Is there an uptime SLA?
Not yet. What is true about availability today:
- Live status: https://www.superdiffs.com/status. It measures the service when you load it; there is no stored uptime or incident history yet.
- We do not offer an uptime SLA yet. Publishing one before we have operated at load would be a number invented for a procurement form.
- A review that cannot complete says so on the pull request and in a check run, rather than posting half a review or nothing. Silence and half a review are the two worst outcomes and both are designed out.
- Security incidents affecting customer data are notified without undue delay, and in any case within 72 hours of confirmation, per the DPA.
How do I report a vulnerability?
Email security@superdiffs.com. Acknowledgement within 48 hours, triage within 5 working days. Machine-readable contact: /.well-known/security.txt.
Safe harbor
Test only against repositories you own. Do not access another customer’s data, do not degrade the service for others, and give us a reasonable window to fix what you find. Stay inside that and we will not pursue legal action for the research.
Can you answer our security questionnaire?
Most security questionnaires are answered by this page plus the security documentation. For the rest — a completed CAIQ, an architecture walkthrough, or the DPA signed — write to security@superdiffs.com and you will get a person, not a portal.
- Terms of Service · Privacy Policy · Data Processing Addendum
- Subprocessor list · Acceptable use
- Benchmark methodology — how we intend to prove the quality claims rather than assert them.