Founding Programme: five places, three months of Growth. Applications close 24 November.Apply for a place

Legal

Security

Last updated 31 October 2026

nineloops holds a company's clients, money, people and conversations, so its security rules live where the data does. This page sets out how your workspace is protected and what you control.

Encryption

Data is encrypted in transit with TLS 1.3, and older TLS versions are refused; browsers are told to use HTTPS only (HSTS, on the preload list). Data is encrypted at rest on our infrastructure. Tokens for the accounts you connect, such as a mailbox or a WhatsApp number, are encrypted with AES-256-GCM under a key held only in the backend secrets. Records in the Docs vault are encrypted the same way, on our server, under a key of your workspace's own that is derived from that secret and bound to your workspace, so a copy of the database or a backup alone does not open them and a value moved to another workspace does not open there. Only members of the workspace can ask for a value to be opened, and the record's own sharing still decides who sees it. nineloops' backend holds the key, so this is not end-to-end encryption. The desktop app talks to the same backend as the web.

Workspaces kept apart

Every record belongs to a workspace, and row level security in the database stops one workspace from reading or changing another's records. The rule runs on the server for every read and every write, however the data is reached.

Access control

  • Eight roles, from owner to guest, and in the apps that give each record an owner (the CRM, Social, Studio, Distribute and Monitoring) a choice of whether people see every record, their team's or only their own.
  • Permissions are enforced by the database, so the app, the API and the AI all get the same answer.
  • Administrators can require two-step verification for every member. The database enforces it on every table, so a session without the second step reads nothing in that workspace, through the app or the API.
  • On Enterprise, people on the company's verified domains sign in through its identity provider with SAML single sign-on, and an administrator can require it. A domain is verified by a DNS record and belongs to one workspace only.
  • Invitations can be limited to the company's own email domains, checked when an invite is made, when an invite link is opened and when a join code is used. An invite link opens only for the address it was sent to, and wrong join codes are counted and stopped.
  • Administrators can ask everyone to sign in again after a set time: every 12 hours, every day, every 7 days or every 30 days. The database holds it too: a session older than that reads nothing in the workspace until the person signs in again. It is off unless an administrator sets it, and nobody is signed out without pressing the button.
  • Each person sees every device signed in to their own account, in Settings, then Sign-in & security, with where and when it was last used, and can sign one out. Each plan sets how many devices one person may use.
  • Five wrong passwords on an account within 15 minutes, and the person it belongs to is emailed.
  • Sensitive HR fields can only be read through an audited call, and pay is visible only to the roles allowed to see it.
  • In Finance, whoever raises a bill cannot approve it, and whoever approves it cannot pay it, whenever someone else can. A company with a single finance manager can still approve and pay its own bills, and the approval records why.
  • Least-privilege access for our own systems and staff.

AI

  • Brain runs on the sign-in of the person using it, so it can only read and change what that person can.
  • Your workspace decides what Brain may do on its own, what it must ask first and what it may never do.
  • Every action it takes is logged with how the records looked before, and most can be undone.
  • AI employees in Brain work inside the permissions you give them, and anything customer-facing waits for a person to approve it.

Audit and your data

  • An audit log of members and settings, and a history on records.
  • Retention rules and legal holds on employee documents in HR, enforced by the database: a document on hold cannot be deleted.
  • An owner or admin can export everything, or one app, from Settings, then Data and audit, whenever they want: a zip with one CSV per table, built from what they are allowed to read, with credentials and tokens stripped. Uploaded files are listed in it by link and are downloaded from the app they live in.
  • Your data lives in Northern California, United States, today, with a move to London planned. Where your data lives sets out every place it goes, including what is sent for AI processing.
  • The owner can delete the whole account from Settings, then Data and audit. There are 30 days to cancel, then everything is removed. Data deletion has the detail.

Infrastructure

We run on established cloud providers, with daily backups. You remain responsible for keeping your own copies of important data. Our pages tell browsers to run only our own scripts (a Content Security Policy), not to be framed by other sites and not to guess file types.

How we test it

  • Every five minutes. The backend checks its six parts: the app, the database, sign-in, file storage, the server functions and scheduled work. Every round is recorded, and that record is the uptime figure shown live on the status page. If a part fails two checks in a row, an alert is emailed to us; if the database fails, at once. The current answer is on our status page.
  • Before a release. The security checks are run by hand as scripts: a probe of the API, a scan of our configuration and a scan of the libraries we use for known weaknesses. The latest API check is on our trust page, with its date.
  • On the live service. The rules that keep companies apart are proved on the live service with throwaway accounts, not only described.

What is independently verified, and what is not yet

  • nineloops itself holds no independent certification yet, such as SOC 2 or ISO 27001, and has not published a third-party penetration test. The tests above are our own, run against the live service.
  • The evidence, live, is on our trust page: the checks on every part of nineloops, the uptime this month and last, and the attack test we ran on our own live service.
  • The services we run on are independently audited: the database and file storage run on Supabase and the app on Vercel, and both hold SOC 2 Type 2 reports.
  • Our uptime record is public: the status page shows, month by month, the share of our checks each part of nineloops answered. The standard plans carry no contractual uptime commitment.
  • Enterprise customers can send us a security questionnaire or review, and we answer it in writing.

No guarantee of absolute security

No product, system or method of transmission or storage can be guaranteed to be completely secure. While we work hard to protect your data, we cannot and do not warrant that the service will be free from unauthorised access, vulnerabilities or loss. You use the service at your own risk, and our responsibilities and liability are limited as set out in our Terms of Service.

Incident response

If we become aware of a security incident affecting your data, we will investigate and tell the workspace owner without undue delay and in any case within 48 hours of becoming aware, as our Data Processing Agreement commits. We also notify the authorities where the law requires. See our Privacy Policy for details.

Responsible disclosure

If you find a vulnerability, email hello@publshr.ae. We will respond promptly, and we ask for a reasonable opportunity to fix it before any public disclosure. Our contact details for this are also published at /.well-known/security.txt.

Start free.

Free for up to five people. Every paid plan includes every app. Browser, Mac and iPhone.