Legal· Last updated 15 October 2026
Sub-processors
Every outside company that stores or processes data on our behalf, what it sees, and where. If a name is not on this list, it does not get your data.
Applies to klaritiq.in and the KlaritiQ platform
The short version
- Four providers see data as a matter of course: Render (hosting, US), Anthropic (AI, US), Resend (email, US) and Cloudflare R2 (uploaded files, Asia-Pacific).
- Google sign-in, Google Calendar, HubSpot CRM import, Jira, Confluence, Microsoft Entra and Slack are optional, and each is switched on by a person who chooses it. The public website loads no third-party tracker: the record of which public pages are viewed, kept only if a visitor accepts the cookie notice, stays in our own database. Public company research also sends company-name queries to Google News; it sends no account or assessment data.
- We will email account holders at least 14 days before adding a provider that would receive personal data.
Always involved
| Provider | What it does for us | What reaches it | Where |
|---|---|---|---|
| Render (Render Services, Inc.) | Runs the application servers and the Postgres database. | Everything: accounts, assessments, documents' metadata, workspace records, logs. | United States, Oregon region |
| Anthropic, PBC | Provides the Claude models that write report prose, propose extractions from uploaded documents, answer Command Center and ask-box questions, draft regulatory mappings and propose public NBFC, manufacturing and diagnostics business observations for human review. | Assessment answers and free text, uploaded document text, chat questions with relevant workspace context, and filtered public company text for active-watchlist research. Recognised personal-contact lines, emails and phone numbers are removed from that public research text. Patient reports and registry contact columns are excluded from public research inference. Not your login details. Anthropic's API terms exclude training on customer data. | United States |
| Resend, Inc. | Sends every email we send. | Your email address, your name if we have it, and the content of the email (sign-in link, report, deletion notice, newsletter). | United States |
| Cloudflare, Inc. (R2) | Holds the raw files uploaded as evidence: policies, minutes, registers, scans. | The file itself, under an opaque key. Not who uploaded it or what was extracted from it; that stays in the database. | Asia-Pacific region (the bucket's location setting) |
Only if you turn it on
| Provider | Triggered by | What reaches it | Where |
|---|---|---|---|
| Google LLC | Choosing "Continue with Google". | Google tells us your name, email and picture; we tell Google nothing beyond the sign-in itself. | Global |
| HubSpot, Inc. | An organisation admin choosing "Connect a CRM". | We read company and deal records once, then revoke our own access. Nothing is written to HubSpot. | United States / EU, per your HubSpot account |
| Google LLC (Calendar) | A member choosing "Connect Google Calendar". | We add, change and remove KlaritiQ's own events on that member's calendar for their open tasks and milestones: the title, initiative name, priority, any required evidence and a link back. We never read their other events. An encrypted refresh token is kept so this can continue. | Global |
| Atlassian Corporation (Jira Cloud) | An organisation admin choosing "Connect Jira" and linking an initiative to a Jira project. | Both ways. We send the title, description, due date, priority, status and owner of the tasks and risks on linked initiatives, a link back to KlaritiQ and, to find the right person, owners' email addresses. We read the same fields back from linked issues and from issues labelled klaritiq, and the project list. Where Jira shows them, we read user email addresses to match people to members. Jira account identifiers and an encrypted refresh token are kept. Every day we report the account identifiers we hold to Atlassian's Personal Data Reporting API and erase any it names as closed or changed. Nothing is deleted in Jira. | Atlassian's cloud, in the region of your Jira site |
| Atlassian Corporation (Confluence Cloud) | An organisation admin choosing "Connect Confluence" and linking a labelled page to an initiative. | One way, read only. We read the pages that carry the label klaritiq and that an admin links: title, space, version history (who changed a page and when) and text. A page with restrictions keeps its history but no text. Confluence account identifiers and an encrypted refresh token are kept. When a site is disconnected, we erase the names and account identifiers of page authors from the recorded history after 3 days, and reconnecting within 3 days prevents that. Nothing is written to or deleted in Confluence. | Atlassian's cloud, in the region of the Confluence site you choose |
| Microsoft Corporation (Entra ID, Microsoft Graph) | An organisation admin choosing "Connect Entra" on the Microsoft Entra permissions card and picking a group for an initiative. | One way, read only. We read the id and display name of the groups an admin picks and the object ids of their members, including members of nested groups, only to count and fingerprint them. The person who connects signs in to Microsoft once so we can confirm the tenant and that they hold an Entra administrator role; we keep only the tenant id from that. A tenant can be connected to one organisation. We keep the group id and name, the role the admin chose, the member count, a SHA-256 hash of the member ids and the tenant id. We never keep member names or email addresses. Nothing is written to Entra. Disconnecting stops the checks and deletes the stored tenant id; the picked groups and the versions already recorded are kept. | Microsoft's cloud, in the region of your Entra tenant |
| Slack Technologies, LLC (Salesforce, Inc.) | An organisation admin choosing "Add to Slack"; each person choosing "Link my Slack account". | The title of any alert KlaritiQ raises about the organisation's work other than the daily brief (for example a task assigned or overdue, a sign-off request, a compliance flag, a regulatory deadline, a risk reopened or evidence expiring, any of which can name an initiative, task or risk), and a link back, plus a short reason if an admin chooses that, which can include a colleague's email address. A person's Slack Home tab shows them their own open tasks and waiting sign-offs. A person's own items go to them as a direct message; urgent alerts go to one channel an admin picks, never one shared with another organisation. Never evidence text or AI-written text. From Slack we receive the workspace, channel names and each linking person's Slack user ID. We read no Slack messages, apart from a single message an organisation admin chooses to save as an evidence note with the Save to KlaritiQ shortcut: its text, its author's name and Slack user ID, the channel name and Slack's ids for the channel and message (and, for someone from another Slack workspace, that workspace's id), when it was sent, a link to it, and the initiative and one-line reason the admin gives are then kept in the organisation's Evidence Locker as an unverified note. To let the admin choose, the names of the organisation's initiatives (up to 99) are shown to them in Slack at that moment. People can also type /klaritiq, and the answer is shown only to them. An encrypted bot token is kept. | Slack's cloud, in the data location of your Slack workspace |
Not sub-processors, but you may wonder
| Who | Why they are not on the list above |
|---|---|
| GitHub | Holds our source code. No customer or personal data lives in the repository. |
| Reserve Bank of India and other regulator websites | We fetch public circulars and RBI releases daily and the RBI company register weekly. Requests include the public URL, our server network address and a research user-agent. No customer account or assessment data is sent. |
| Google News (Google LLC) | Free RSS searches receive public company names and sector keywords, our server network address and a research user-agent. They do not receive assessment answers, rep-entered contacts or customer account details. No paid news API is connected. |
| Company websites | We read public company homepages, footer links, grievance and privacy pages on a quarterly cycle or when a lead needs a new check. These servers receive the public page URL, our network address and a research user-agent. Officer names/contact details are not saved as observations, and raw pages are not retained. |
| Apple public app feeds | When an Apple app ID is configured for company research, Apple receives that public app ID, its country, our network address and a research user-agent. We read listing descriptions and turn at most 50 recent review entries into deterministic theme counts. Raw review text and reviewer identities are neither retained nor sent to Claude. Google Play is not connected. |
| Configured public publishers and company/vendor pages | Curated public HTML/PDF URLs may be fetched for CRISIL, ICRA, CARE, India Ratings, CERT-In, BSE/NSE, company careers pages and vendor customer pages. Publishers receive the public URL, our network address and a research user-agent. Filtered business text can go to Claude for active-company proposals, all pending human review; raw documents are not retained. Job-board scraping and paid document APIs are disabled. |
| NABL/QCI, ABDM/NHPR, CDSCO and public health-scheme registers | Configured public medical-laboratory, facility, device, clinical-establishment and PMJAY/insurer empanelment exports support corporate directory checks. These websites receive the public URL, our network address and research user-agent. Only corporate identifiers, city, scope and dated membership evidence are retained; no patient records or individual contact columns. Unreadable or incomplete lists cannot establish absence. |
| SEBI, ministry PLI portals, FDA and industry/partner publishers | Configured public offer filings, PLI beneficiary lists, FDA warning letters or published inspection observations, industry programme lists and corporate portfolio/client/member pages support manufacturing and diagnostics research. They receive the public page URL, our network address and research user-agent, with no account or assessment data. Filtered business passages can go to Claude for active-company proposals; public registry rows are parsed deterministically and raw documents are not retained. |
| The AI assistant you connect (Claude, ChatGPT, Copilot or similar) | An owner turns on AI assistants and a person connects one with their own sign-in, so the assistant's provider is your choice, not ours. KlaritiQ sends nothing on its own: the assistant asks and we answer as that person, read only. What it reads goes to its provider under your agreement with them. Every request is recorded (who, which tool, how many rows), never the answer. |
| Your own HRIS, Active Directory or webhook endpoint | These are your systems, connected on your instruction; data flows from them to us, or from us to a URL you chose. |
No paid MCA document vendor or paid news/review API is connected to Leads. Curated publisher URLs, complete sector-directory exports and Apple app IDs must be configured before their scheduled fetches run; the admin coverage panel reports missing configuration and degraded fetches. Public official landing pages may be checked manually to verify availability. Adding a fetcher changes this page and the Privacy Notice in the same release. Paid document access also requires a displayed price and a rep's acceptance for a qualified lead that has reached contacted stage.
Two locations, on purpose
Everything in the database is in the United States (Render, Oregon). Uploaded files are the exception: the R2 bucket that holds them is set to Cloudflare's Asia-Pacific region, so a document you upload does not leave the region. If your organisation needs the database in India as well, see the residency note on the Privacy Notice.Changes to this list
Adding a provider that would receive personal data: email to account holders at least 14 days before, with the right to delete your account before the change. Removing one, or a change that affects nobody's data: the date at the top of this page moves and nothing else happens.
Questions about anything on this page: hello@getklaritiq.com. We write these pages ourselves, in plain language, to match what the product actually does; they are not a substitute for legal advice to you.
Talk to us
Questions about this page?
Write to the KlaritiQ team about anything on it, including privacy requests and complaints.
Email the support team