Data Processing Agreement
Version 2026-11-07.
This Data Processing Agreement (“DPA”) forms part of, and is subject to, the Terms of Service between you (“Customer”) and TOTEMDEV SRL (“WikiFix”, “we”) for WikiFix at app.wikifix.ai and WikiFix for Confluence on the Atlassian Marketplace (together, the “Service”). It applies whenever we process personal data on your behalf and the GDPR or UK GDPR applies to that processing. No signature is required – accepting the Terms of Service accepts this DPA. If your procurement process needs a countersigned copy, email security@wikifix.ai and we’ll send one.
You are the controller of the content of the sources you connect (Confluence Cloud sites and Notion workspaces). We are your processor: we process it only on your instructions, to provide the Service. We don’t use it to train AI models, or to tune or improve the AI prompts behind the Service.
When a person reads your content. We read your content only when you ask us to in writing or give us explicit written permission, such as in a support email, and only for the problem you raised, unless the law requires otherwise. We keep that request. When we read your content this way, we act as your processor. A prompt we change after looking into a problem holds none of your text.
Our own records. For your account, billing, and usage of the Service, we are an independent controller, as our Privacy Policy describes. That covers sign-in and account data, usage events, session replay, error reports, server logs and the feedback a person chooses to send us. Usage events, error reports and server logs identify people by our internal ids, and hold no page content. Server logs also hold the IP address of each request, for security. Sign-in and billing keep the email address they need. Feedback holds only what a person chooses to write to us, and their email if they leave “Reply to me” ticked. We add no page content to it. None of it is listed below as personal data we process for you. It appears in this DPA only where deleting your team doesn’t reach it, under How long we keep it.
The terms themselves
The operative clauses of this DPA are the Common Paper DPA Standard Terms, Version 1.1, an attorney-drafted open standard published under CC BY 4.0, incorporated here by reference with the changes under Changes to the Standard Terms. Where those terms refer to the Cover Page, this document supplies it. Where this document and those terms differ, this document applies, except that the SCCs prevail over both. The defined terms are:
| Cover Page term | Value |
|---|---|
| Governing Member State | Ireland – the EEA SCCs are governed by Irish law (Clause 17) and disputes under them are heard in the Irish courts (Clause 18(b)) |
| Competent supervisory authority | The authority of the EU member state where you are established. If you are not established in the EU, the authority of the member state where the people whose data is processed are located |
| Approved Subprocessors | The list published on our Sub-processors page, giving each sub-processor’s identity, country and processing tasks |
| Security Policy | How we protect it below and our Security page |
| Report | None. We hold no third-party audit report today – see Audits below |
| Provider Security Contact | security@wikifix.ai |
| Special Category Data | Only on the Restricted or Strict Data handling level – see Special categories below |
| Special Category Data Restrictions or Safeguards | As set out under Special categories below |
Changes to the Standard Terms
| Standard Terms | Change |
|---|---|
| Section 5.1 | The Standard Terms keep compliance records for 3 years after the DPA ends. The change history is the exception: each entry is kept for three years from its date, as stated under Data handling. |
| Section 5.2 | Section 5.2 of the Standard Terms does not apply. It says we are regularly audited by independent third parties and can send you a report; we aren’t and we can’t yet. |
| Section 8.3 | The second sentence, on liability between the parties under the SCCs and UK Addendum, is replaced by Liability below. |
Liability
To the maximum extent the law allows, each party’s total liability arising out of or related to this DPA and the SCCs, including the UK Addendum, in contract, tort or on any other theory, is subject to the exclusions and the cap on liability in the Terms of Service.
This does not limit what an individual can claim for their own data-protection rights under the GDPR, the UK GDPR, the SCCs or the UK Addendum, or any other liability the law doesn’t let a contract limit. Where the SCCs don’t allow a limit in this section, the SCCs prevail.
Parties
| Provider (data importer, processor) | TOTEMDEV SRL, company number (IDNO) 1022609000314, Miron Costin 19/4, ap. 37, Chișinău MD-2068, Republic of Moldova |
| Customer (data exporter, controller) | The team that signs up at app.wikifix.ai, or the Atlassian organization that installs WikiFix for Confluence. Its name, address and contact person are those in its account, or those it gives us on request |
| Contact | Rezgar Cadro, founder, security@wikifix.ai |
What we process, and why
Data subjects. Your Confluence and Notion users – page creators, authors and owners, people named in page content or comments, and whoever starts a scan or applies a fix.
Categories of personal data. Only what we process for you:
- Confluence and Notion page content within the spaces and pages you scope to a scan. This is your text, and it may contain personal data of any kind your users have written into it. We don’t select what it contains.
- What we derive and store from that content – the text listed under How long we keep it, which can repeat personal data the pages contain.
- Directory ids – the Atlassian or Notion account ids of page creators, owners and last editors, display names and group membership as the source returns them, and whether an account is still active.
- The notification address – the address a scan notification goes to. In WikiFix for Confluence it is resolved from Atlassian at send time and not stored. In WikiFix at app.wikifix.ai it is the address on the member’s account, which is stored encrypted.
Everything else we hold about the people who use the Service is ours, as controller, and is in our Privacy Policy: account and sign-in data, billing, usage events, error reports, server logs and feedback.
Special categories. This means data revealing racial or ethnic origin, political opinions, religious or philosophical beliefs or trade-union membership; genetic, biometric, health, sex-life or sexual-orientation data; and criminal-offence data. We don’t look at your content, so we can’t tell whether it holds any. If it does, you must use the Restricted or Strict Data handling level. If you use Standard with such data anyway, that is your decision and your responsibility.
You also control what we read: a scan only ever reads the content of the spaces and pages you scope it to. We never read the content of a space or page you leave out of scope. To show you what you can choose, and for Notion to place a page in its workspace, we do receive the names, ids and structure of what your account can see or what is shared with the integration. That stays in memory for minutes and isn’t stored; we store only the id of each space or Notion page listed.
Safeguards for sensitive data (SCC clause 8.7). Because we can’t rule such data out, these apply to everything we process:
- Limited access. One person operates WikiFix and administers production, under a duty of confidentiality. Your requests reach only your own team’s rows. The key that decrypts stored text is granted only to a named list of roles of our running Service. The person who administers our cloud account could also reach it.
- Reads only on your instruction. We read your content only as set out at the top of this DPA, and we keep your request or permission. We don’t keep a technical log of each read, in our tools or in a vendor’s own console, such as Langfuse’s.
- Strict purpose. We process your content only to provide the Service, on your instructions. We don’t use it to train AI models, or to tune or improve prompts, and a prompt we change after looking into a problem holds none of your text.
- Also: scope confined to the spaces and pages you choose; storing what is listed under How long we keep it rather than a copy of your wiki; encryption in transit, in our AWS account at rest, and column by column for the text we store; processing in the EU (all of it with EU residency required; without it, all but two analysis checks, which run at TypeSafe in the US with zero data retention; see Data handling); and deletion on the schedule below.
Purpose. Reading pages in the spaces you scope; deriving an index from them; sending page text to AI models to find contradictions and author fix text; on the Standard level, recording those AI requests and responses so that a problem can be looked into when you ask us to; storing findings and the text that evidences and applies them; writing fixes you approve; posting a comment when you ask a page owner to decide; and sending scan notifications.
Frequency. Continuous for as long as you use the Service – on each scheduled or on-demand scan, and each time you apply a fix.
Duration. For as long as you use the Service, which ends when you delete your team or uninstall WikiFix, or when you ask us in writing to end it. Then as set out below.
Data handling
Your team chooses a level in WikiFix under Admin → Data handling. During onboarding, the first admin to answer is asked about sensitive data and EU residency, and later admins aren’t asked again. It starts on Standard. Choosing a level is your instruction to us under the Standard Terms.
Whatever the level, we keep what WikiFix needs to work: the database, account and billing records, and server logs, listed under How long we keep it. The level decides everything else:
| Standard | Restricted | Strict | |
|---|---|---|---|
| AI requests and responses, at Langfuse | Full text, about 30 days | Cost and timing with our internal ids, and no text | Not recorded |
| Two analysis checks, at TypeSafe in the US | Only without EU residency | Only without EU residency | Only without EU residency |
| The same two checks, at eu/jev in Germany | Only with EU residency | Only with EU residency | Only with EU residency |
| Error reports and performance traces, at Sentry | Sent, with no page content | Sent, with no page content, site address, email, name or Atlassian or Notion account id | Not sent |
| Usage events, at PostHog | Sent, with no page content | Sent, with no page content, site address, email, name or Atlassian or Notion account id | Not sent |
| Session replay, at PostHog | Recorded, with text and inputs masked, and elements that hold your text in an attribute left out | Recorded, with text and inputs masked, and elements that hold your text in an attribute left out | Not recorded |
| Feedback a person writes to us, at PostHog | Sent if they send it, with their email if they leave “Reply to me” ticked | Sent if they send it, without their email | Sent if they send it, without their email |
| Report and space names in scan emails, at Brevo | Sent | Not sent: counts and links only | Not sent: counts and links only |
The web sign-in and sign-up screens come before we know a person’s team, so on every level they send error reports and a page-load event as on Standard. These hold no page content, and they aren’t tied to your team. PostHog is set to discard IP addresses, and these screens are not recorded.
- Cost. Neither the level nor EU residency changes what a scan costs.
- Looking into problems. Without AI records, we can’t look into past AI results if you report a problem.
- Moving to Restricted or Strict. We delete your team’s records at Langfuse, including those from a scan still running when you moved, usually within about a day, and longer for a team with many records. A scan already running finishes the step it is on at the old level. Moving back to Standard records new requests only.
- Two analysis checks. They help decide whether a finding stands. With EU residency required, on every level, they run on different AI models, which reside in the EU: eu/jev, operated by Bevelites GmbH on its own server in Germany. Every other AI step stays on AWS Bedrock in the EU. Without it, on every level, they run at TypeSafe in the US. Either way, a request carries the passages in question, which can come from several pages, each with its page’s title and headings and the section around it, up to 2,500 characters of section per passage. Neither receives a whole space. eu/jev doesn’t store the excerpts it receives, in its database or its logs, and doesn’t use them to train AI models. We use TypeSafe with zero data retention: it doesn’t keep the excerpts it receives or use them for anything else, including training AI models.
- WikiFix for Confluence only: the WikiFix agent in Rovo. When a person uses it, Atlassian’s Rovo AI receives finding text (the questions and answers), page titles and the passages behind them, the text of any page the agent reads for them, and, when they settle a finding, the before and after text. Your page bodies are already at Atlassian. This is Atlassian’s service, under your own Atlassian agreement, and it is the same on every level, Strict included: the level doesn’t change what it receives. To settle a finding the agent keeps one summary per person in Atlassian’s Forge storage, with the question, page title and before and after text, until that person settles it or prepares another. We set no expiry on it while WikiFix is installed, and our deletion job doesn’t clear it. Atlassian deletes app storage after an uninstall, on its own retention schedule.
- Change history. Admin → Change history shows each change to your configuration, permissions and consent: who, when and what. That covers your level and residency, your own AI key (never the key itself) and sources, who may run scans, members and invitations, onboarding and the acceptance of these terms, and share links created, revoked and opened. It holds no page content. We keep each entry for three years from its date, even after your team is deleted, so you can show what applied when. To tie it together, a deleted team keeps our ids for the team and its people, and their Atlassian site, account and group ids and Notion workspace ids, and nothing else about them. To have all of it deleted, email security@wikifix.ai and we will.
How long we keep it
We read pages live from your source on each scan and don’t keep a copy of your wiki. What we do keep:
| What | Where | Until |
|---|---|---|
| Text derived from your pages, which can repeat their wording and is kept whether or not it leads to a finding, and a numeric index of them. For each fix previewed, the text it would replace and its proposed text, so it can be applied and reverted; for a merge of duplicate pages, the whole previous body of each page merged away. Questions and answers on a finding, comments you ask us to post, and report names. A search index of your findings. Scan history, ids of page creators, owners and last editors, and account records. See How we protect it for what is encrypted | Our database in AWS Frankfurt | About 30 days after you uninstall WikiFix for Confluence or delete your team: a daily job starts the deletion on its first run past the 30-day mark, once every install on your team is uninstalled. Disconnecting a source deletes the sign-in tokens we hold for it, and nothing already stored from it, which stays until the team is deleted. Deleting a team at app.wikifix.ai deletes its tokens at once. An uninstall stops scans of that site at once and deletes the app token we cached for it; Notion workspaces and other Confluence sites added by hand stay connected for the 30 days, with their scheduled scans waiting, and their tokens are deleted when the 30 days are up. When someone is removed from a team, we stop using their sign-ins and ask Notion to revoke theirs. We ask Notion to revoke the Notion tokens in each case; Atlassian offers no way to revoke a token, so a person’s grant stays listed in their Atlassian account until they remove it. Database backups roll off on a rolling 7-day window. We keep no long-lived snapshots |
| Sign-ins: at app.wikifix.ai, an email address and password verifier; for WikiFix for Confluence, a record named with the person’s Atlassian account id, with no email address or password | Amazon Cognito in AWS Frankfurt | With the team’s data as above, unless the person still belongs to another team |
| On the Standard level, the full text of each AI request and response, which embeds the page text under review | Langfuse | About 30 days after it was recorded, by a separate daily job, or usually within about a day of moving to Restricted or Strict, and longer for a team with many records |
| Without EU residency: page excerpts sent to TypeSafe as part of generating findings | TypeSafe, in the United States | Not kept: we use TypeSafe with zero data retention, so it doesn’t keep the excerpts it receives or use them for anything else, including training AI models |
| With EU residency: page excerpts sent to eu/jev as part of generating findings | eu/jev (Bevelites GmbH), in Germany | Not stored: it doesn’t store the excerpts it receives, in its database or its logs, and doesn’t use them to train AI models |
| The emails we send you and their delivery records, as the Sub-processors page lists what they hold | Brevo | Brevo’s terms set no fixed period |
| Short-lived records of a sign-in in progress: a nonce with the person’s Atlassian account id and site id (about 5 minutes), and, while an admin picks a site to connect, the sign-in state with encrypted tokens (10 minutes) | DynamoDB in AWS Frankfurt | Deleted automatically when they expire, which AWS can take up to about 48 hours to do; the nonce table’s point-in-time backup keeps removed items for up to 35 days |
| WikiFix for Confluence only: the summary the Rovo agent keeps to settle a finding for one person | Atlassian Forge storage | Until that person settles it or prepares another. No expiry while WikiFix is installed. Our deletion job doesn’t clear it; Atlassian deletes app storage after an uninstall, on its own retention schedule |
| Your change history, with no page content, and the ids that tie it together: ours, Atlassian site, account and group ids, and Notion workspace ids | Our database in AWS Frankfurt | Three years from each entry’s date, even if your team is deleted, unless you ask us at security@wikifix.ai to delete it sooner |
| A numeric index with our internal ids, and no page text | Neo4j Aura, in AWS Frankfurt | The end of the scan run; 24 hours at most if cleanup fails |
| Server logs, with IP addresses, internal ids and our vendors’ reference numbers. No page content | Amazon CloudWatch in AWS Frankfurt | 90 days |
| Database performance records | Amazon RDS Performance Insights | 15 months |
| Analytics events and session replays | PostHog, stored in the EU | Events 3 months, deleted at PostHog’s next monthly cleanup; session replays 30 days |
| Error reports and performance traces | Sentry, stored in the EU | 30 days |
Deleting your team doesn’t reach any row from Langfuse down. Each expires on its own clock, apart from the Rovo summary, which has no expiry from us while WikiFix is installed.
When the Agreement ends, we delete your data, unless you ask us to return it instead. You can also ask us to delete it sooner. We do both by hand.
How we protect it
| Measure | What we do |
|---|---|
| Encryption | TLS 1.2 or higher in transit. At rest in our AWS account, the database and its backups are encrypted with AES-256 under a KMS key we manage; storage buckets, secrets and the short-lived sign-in tables use AWS’s own managed encryption; log groups use CloudWatch’s default encryption, with no key of ours. Text we store in the database (text derived from your pages, fix text and previous page bodies, questions, answers and comments) is also encrypted column by column, and search doesn’t read your text. We hold the key, and the running Service can decrypt with it. The numeric index, keyed hashes used for search and matching, report, source and team names, and your site address are not encrypted column by column. Vendors that hold content use their own encryption |
| Credentials | Your Atlassian tokens and any AI key you bring are envelope-encrypted (KMS data key plus local AES-256-GCM) with automatic rotation. Our API never returns them, plaintext travels only over TLS to the vendor it belongs to, and we don’t log them by design |
| Tenant isolation | Customer requests run as a database role that cannot bypass PostgreSQL row-level security, pinned to their team. Every table that holds team data has that protection. The one table without it lists the teams themselves, and our code filters it. A few named system jobs (such as sign-in lookup, token rotation and scan scheduling) read specific tables across teams, and the operator’s admin credential bypasses row-level security, though it sees encrypted columns only as ciphertext |
| Access control | One person operates WikiFix and administers production. Access to our cloud account is through single sign-on; deployments run from our source repository through a deploy role. The database is Amazon Aurora inside that account, encrypted at rest, with generated passwords held in AWS Secrets Manager, which only identities in that account can read |
| Audit logging | Administrative actions in our production account in Frankfurt are recorded to a tamper-evident audit log – written continuously, integrity-validated, retained for seven years, with a searchable one-year copy |
| Region | Storage and compute in AWS Frankfurt (eu-central-1); default AI inference on AWS Bedrock through EU inference profiles, which can serve a request from any EU region, with cross-continent profiles refused for our analysis models. Every sub-processor that offers an EU region is pinned to it, except Atlassian for contact-form submissions from our website. Where an install requires EU residency, all AI processing runs in the EU, with two analysis checks at a sub-processor in Germany. Where it doesn’t, two analysis checks go, on every level, to the one sub-processor located in the United States for your content, named on the Sub-processors page, which keeps none of what it receives and uses none of it for anything else, including training AI models. Where an install declares a residency requirement, processing that cannot meet it is refused rather than performed elsewhere, unless you have also added your own AI key, which sends AI calls to that AI provider whatever residency you have set |
| AI | Default inference runs on AWS Bedrock; per Bedrock’s terms your prompts and completions are not used to train base models and the model provider has no access to them |
| Telemetry | Server logs, error reports and analytics hold no page content, titles or text derived from them on any level. Feedback holds only what a person chooses to write to us, and we add no page content to it. Backend analytics events pass a per-event allow-list and browser events a block-list of content and external-id fields. Session replay masks text and inputs, leaves out elements that hold your text in an attribute, and records layout and clicks. Error reports carry stack traces, error codes and page addresses, with no names or emails. Performance traces carry the timings of requests and page loads, with addresses but no query strings, and none are sent on Strict. Data handling sets what else each carries |
The Security page has more detail.
Other commitments
- Confidentiality. Everyone authorised to process your data is bound by a duty of confidentiality.
- Unlawful instructions. If we believe an instruction from you breaches data-protection law, we will tell you rather than quietly carry it out.
Breach notification
If we become aware of a personal data breach affecting your data, we will notify you without undue delay, and no later than 72 hours after becoming aware – with what we know at the time, and updates as we learn more. We’ll also take prompt steps to contain and investigate it.
Your own 72-hour deadline under Art. 33 runs from the point we tell you, not from when the incident happened.
Audits
We hold no SOC 2 or ISO 27001 certification today. That is why the Report field above reads “none” and Section 5.2 of the Standard Terms is disapplied.
In its place:
- Security due diligence. We answer information-security, due-diligence and audit questionnaires in writing, and provide documentation of the measures above. Send them to the Provider Security Contact; the Standard Terms set the cadence at once a year.
- Audit. We will give you the information reasonably necessary to demonstrate compliance with this DPA, and will contribute to audits as the Standard Terms provide. Your audit rights under SCC clause 8.9 apply and prevail. Where a supervisory authority or applicable law requires more, we will meet that requirement.
- Records. We keep records of our compliance with this DPA for three years after it ends. The change history is kept as stated under Data handling.
Sub-processors
We use the sub-processors listed on our Sub-processors page, each with its purpose, the data it receives, and its location. That page is the authoritative list, and this DPA constitutes your general written authorisation for them under Art. 28(2).
Each sub-processor processes your data only for the role on the Sub-processors page, for as long as you use the Service, and then as set out under How long we keep it.
We publish a change there at least 10 business days before a new or replacement sub-processor begins processing your data. Every change is published to our sub-processor feed at the same moment – follow it and you’ll be told without having to watch the page.
You then have 30 days from publication to object on reasonable data-protection grounds. If you do, we’ll work with you in good faith to resolve it; if we can’t, you can stop using the Service at any time, and we delete or return your data as this DPA provides. Email security@wikifix.ai to raise an objection.
International transfers
Your content is stored in the EU and, where your install declares EU residency, our AI processing runs only there, including two analysis checks that a sub-processor performs in Germany, unless you have also added your own AI key or use the Rovo agent. Where no residency is declared, on any Data handling level, two analysis checks are performed by a sub-processor located in the United States, which keeps none of what it receives and uses none of it for anything else, including training AI models, as set out on the Sub-processors page. We are, however, established in the Republic of Moldova, which is not covered by an EU adequacy decision – and our administering the Service from Moldova is itself a transfer under Chapter V of the GDPR. We’d rather state that plainly than leave you to discover it.
That transfer is covered by the EU Standard Contractual Clauses (Commission Implementing Decision 2021/914), Module 2, controller to processor, incorporated into this DPA by reference. You are the data exporter and we are the data importer; the sections above supply the information required by SCC Annexes I and II, and the Sub-processors page supplies Annex III. For customers subject to the UK GDPR, the UK Addendum to the SCCs applies on the same basis.
The Clauses are governed by the law of Ireland, and disputes under them go to the Irish courts – the SCCs require an EU member state for both, so this sits alongside rather than replaces the governing law of our Terms. Nothing in that choice narrows your rights: under the Clauses, a data subject may also bring proceedings in the courts of the member state where they live.
The Clauses require a transfer impact assessment (clause 14), and we keep a written one for this transfer. Alongside the clauses, these facts are relevant to your own: your data is stored at rest in the EU, and nothing is stored in Moldova, apart from working files from a read you permitted in writing, which we delete afterwards; that lowers the risk but doesn’t remove the transfer; it is encrypted in transit and at rest; Moldova is a party to Council of Europe Convention 108 and has ratified the protocol that modernises it (Convention 108+), which is not yet in force; and Moldovan Law No. 195/2024, in force since 23 August 2026, transposes the GDPR into national law.
Questions
security@wikifix.ai – a person answers.