Security

Your data is yours, and kept apart

What we do to keep your company's information safe on hosted IMS Creator, in plain terms. It summarises our architecture and our self-assessment against the OWASP Application Security Verification Standard, level 2.

Kept apart from every other company

  • Each company has its own databases, one per site and one for the company, and its own place in the file store.
  • No query or screen spans companies. On every release a test signs in as one company and asks for every page with another company's names and ids, and must get nothing back.
  • People belong to one company and sign in with an address at one of your approved email domains, which you verify.

Our operators cannot see your content

  • Platform administration shows companies, sites, how many people, and the service's health. Never a document, a file name, a person's name or what a job is about.
  • It cannot open your sites. Our administrators sign in with an authenticator, and from outside only through Cloudflare Access as themselves.
  • If we sign your people out in an incident, your company admins see it in your audit log.

Signing in

  • An authenticator code is asked for before the password, so a password leaked elsewhere cannot even be tried. Your company can require it for everyone.
  • Passwords of at least 12 characters, checked against known breached passwords without sending them anywhere, and stored with Argon2id.
  • Repeated wrong attempts lock the account for a while and email the person. Resets are by single-use links and end every other session.
  • A hash-chained audit log of sign-ins and changes, so an entry changed afterwards shows. It can be exported.

Your drawings are not stored

  • Drawings are read from a folder on your computer, processed in batches, and written back to a folder of yours.
  • On our server they exist only while a batch is processed, on an encrypted disk that is not backed up, and are deleted when you collect the results, or within 2 hours at most.
  • What we keep for the register: each drawing's fingerprint, name, title block details, tags, links, summary and small page images.

Claude, on your own account

  • Drafting runs on your own Claude account: the Anthropic API, AWS Bedrock or Google Vertex AI, in the region you choose. For processing in Australia, Bedrock or Vertex in an Australian region.
  • Your key is checked when entered, never shown again, encrypted with your company's own key, and used only for your work.
  • If your key fails your Claude work pauses and your admins are told; unusual use is emailed to them.

Infrastructure in Australia

  • A dedicated, encrypted virtual machine in Australia behind Cloudflare: no port is open to the internet, with Cloudflare's firewall, bot protection and rate limits in front.
  • All traffic encrypted (TLS 1.2 or later, HSTS). Security fixes applied automatically; administration through a private network with keys only.
  • All customers' data is held in Australia. Your Claude processing is wherever your provider and region are.

Backups and recovery

  • Databases archived continuously, with at most 15 minutes of changes at risk, and nightly backups of each database on its own, so your company can be restored without touching anyone else's.
  • Encrypted before they leave the machine, kept on a separate backup server and copied off-site: daily for 30 days, monthly for 12 months.
  • A restore tested every quarter, against a target of 4 hours.

Plugins in a sandbox

  • Plugin functions run as WebAssembly with no files, no network and limits on memory, processing and time.
  • Each runs inside one site, with only the permissions your admin consented to, and only if signed with the platform's key after review.
  • Installs, updates and removals are in your company's audit log, and a plugin can be suspended everywhere at once.

If something goes wrong

  • We act at once to contain it: we can suspend a company, end sessions, rotate keys, or take the service offline. We preserve the evidence.
  • We tell affected customers within 72 hours of becoming aware, with what we know and what you should do, and help you with your own obligations.
  • Next, before we grow further: an independent penetration test (then yearly, with its summary shared with you), keys in a cloud key vault, central logging and signed images.

Reviewing us for your organisation?

Read our security pack: the architecture and data flows, what is sent to Claude and what your provider keeps, access, operations, incident response, our service providers, answers to common questionnaire questions and the ASVS level 2 self-assessment. To report a security concern, write to [email protected].