Security

Last updated October 2026

Where we stand

FilliQ handles identity numbers, bank details and signed documents for businesses, so this page says what is built, in plain terms, and what is not.

FilliQ is not certified to ISO 27001 or SOC 2, has not been audited against either, and has not had an independent penetration test. Nothing on this page is an attestation or a certification. It describes what the software does, and its main claims are held to the code by automated tests in the product's own repository. Our providers (below) publish their own certifications and reports.

A customer that needs more than this page can ask for our vendor security questionnaire, which answers common due-diligence questions one by one and says "No" or "Not yet" where that is the truth.

Encryption

  • Everything travels over HTTPS, and browsers are told to insist on it (HSTS, two years).
  • Identity numbers, passport numbers, and bank and card-like values are encrypted a second time before they are stored, with AES-256-GCM. The key for each organisation is derived separately from a master secret that is not kept in the database, and the organisation and the field are bound into the encryption, so a value copied from one organisation's row into another's does not decrypt. A copy of the database alone does not reveal them. If the secret is missing the application refuses to store the value rather than store it in the clear.
  • An organisation's own Anthropic key and the signing secret of its audit stream are kept the same way, and are never shown again after they are set.
  • API keys and assistant connections are stored as one-way hashes. The secret is shown once, at creation.
  • Uploaded forms, scans and completed documents are kept in private storage and reached only through short-lived signed links.
  • The database provider also encrypts its disks and backups. That layer is theirs.

Access control

  • Each organisation's data is separated in the database by row-level security on every table: a signed-in person reads only the rows their organisation, team, folder and role allow, and the database enforces it, not only the application.
  • Roles: owner, admin and member on every plan with an organisation; user admin and billing admin from Business, which run the organisation without opening its forms; and custom roles on Enterprise that choose exactly what each person or team may do (fill, train, export, send client links, approve, bulk fill, view the access log, manage members, billing or security).
  • Forms can be put in folders and handed to teams, so a person sees the forms they have been given.
  • Programs and AI assistants get scoped keys or an approved sign-in that can do only what the scope says, never read the access log or manage members unless that scope was granted, and are limited in how fast they can call.
  • The powerful credentials stay on the server. The code that uses the database's service role is limited to a short list of named reasons, and a test fails if anything else reaches for it.
  • A client opening an intake link must enter a code emailed to them, with a limit on attempts.

Sign-in, single sign-on and provisioning

  • Business and Enterprise organisations can require two-factor sign-in (an authenticator app) for everyone, and limit how long a sign-in lasts. The database refuses a session that does not meet the organisation's rule, whatever sends it.
  • Business and Enterprise organisations can use SAML single sign-on with their own identity provider, on domains they have proved they control.
  • Enterprise organisations can provision and remove people from their identity provider with SCIM.

Audit log

  • Every time a form is opened, filled, exported, signed or sent, who did it and when is recorded. The record holds no field values by design: the database refuses an entry that carries an answer or a value.
  • Business and Enterprise organisations can read the log and download it as a CSV; each download is itself recorded.
  • Enterprise organisations can have the log sent, as it happens, to their own HTTPS endpoint. Each delivery is signed so the receiver can check it came from FilliQ, and the address is checked so the feature cannot be pointed at a private network.

Signatures

A person can sign a form in FilliQ, and a client can sign through an intake link. Each signing records who signed, when, and from which device, and a finished document can be sealed: its fingerprint (SHA-256) is stored, so a later change to the file is detectable. FilliQ never fills a signature, witness or certification field on its own, and never signs for anyone. These are not qualified or certified electronic signatures.

Retention and deletion

The organisation sets how long answers, uploads and completed forms are kept. A scheduled job deletes what is due, removes the stored files as well as the rows, and checks afterwards that both are gone. Remembering a returning client's details is off unless the organisation turns it on. A person can delete their account, and an organisation can remove a member and delete their data. The retention periods and the exceptions (a legal hold, a provider's backups) are in the Privacy Policy.

AI

  • FilliQ uses Anthropic's API to read forms and suggest which saved detail answers which field. It does not train models on customer data and does not use one customer's data for another.
  • The AI never submits a form to anyone, and never fills a signature, a witness or a certification field. Fields it is unsure of are left empty, and a person reviews every form.
  • An Enterprise organisation can use its own Anthropic key. Its AI requests are then made on its own account, under its own agreement with Anthropic, and if that key is refused FilliQ reports the failure and does not fall back to FilliQ's key.
  • AI can be switched off for a deployment altogether.

Backups and availability

The database is hosted by Supabase, which takes the backups; FilliQ does not hold separate copies and does not publish a recovery time or recovery point. Deleted data can remain in the provider's backups until they roll off. There is no uptime commitment, and a service-level agreement is not offered at present.

How the software is built

  • Every change runs a large automated test suite, against a real PostgreSQL database for the rules that live there, before it is deployed. The deployment is stopped if the database does not match what the code expects.
  • The database changes only through numbered migrations that are recorded and checked for tampering, and a check after each deploy confirms that row-level security is on for every table.
  • A scan refuses a build that contains a credential.

Reporting a security problem

Write to support@filliq.co with "Security" in the subject, say what you found and how to reproduce it, and please give us a chance to fix it before you share it more widely. Do not include real identity numbers or customer data in the report. We do not run a paid bounty programme.

Providers and more detail

The companies that process data for FilliQ are listed in the Privacy Policy, with what each receives. A customer that needs its own database, hosting and keys can have FilliQ run as a private instance, by arrangement. The Privacy Policy covers what is held, for how long, and how to ask for it to be corrected or deleted.

See also the Privacy Policy. Security reports: support@filliq.co