Security pack

Common questions

Version 1.0, November 2026 · The whole pack · PDF

Short answers to what security questionnaires ask most. Each links to the section with the detail.

The company and the service

Question Answer
What is the service? A web application for writing and running an integrated management system (ISO 9001, 14001, 45001, 55001 and others): its documents, registers and plans, at one or more sites.
Who runs it? IMS Creator, in Australia.
Where is our data held? In Australia: the server, its backup server and the off-site copy. Claude processing is where your own Claude provider and region are. Data
Is it multi-tenant? Yes, with a database per company and per site, and a separate place in the file store per company. Data
Do you have ISO 27001 or SOC 2? Not yet. An OWASP ASVS level 2 self-assessment is published; an independent penetration test comes before the service grows beyond its first customers, then yearly. Overview
Can we see a penetration test report? When the first one is done, its summary is shared with customers.

Data

Question Answer
Is data encrypted at rest? Yes: the data disk (LUKS2), your secrets again with your company's own key, and backups encrypted before they leave the server. Data
In transit? Yes: TLS 1.2 or later everywhere, with HSTS. Architecture
Can your staff read our documents? Our console cannot show them. Server access is limited to the owner, over a private network, for operating the service, such as a restore you ask for. Data
How long do you keep our data? While you subscribe. Drawings themselves: 2 hours at most. Logs: 90 days. Backups: 30 daily and 12 monthly. Data
Can we get our data out? Documents as Word and registers as Excel at any time; a full export on request and when you leave. Data
What happens when we leave? A full export, then deletion of your databases, files and keys; backups expire within 12 months; confirmed in writing. Data
Is our data used to train AI? No. Not by us, and under Anthropic's Commercial Terms not by Anthropic; AWS Bedrock and Google Vertex AI make the same commitment. Claude

Claude and AI

Question Answer
Whose AI account is used? Yours: Anthropic API, AWS Bedrock or Google Vertex AI, under your own agreement. Claude
What is sent to the AI? Only what each piece of work needs: your site's profile and related documents to write a document; page images to read drawings; asset classes and a work history summary for asset information. Claude
What does the provider keep? Anthropic: not retained by default except for its covered models (30 days); longer only if flagged or required by law; zero data retention available. Bedrock: not stored. Claude
Can we keep AI processing in Australia? Yes: AWS Bedrock or Google Vertex AI in an Australian region. Claude
Does the AI act on its own? No. Its answers are proposals that people review and approve; it has no tools that act on the service or your systems. Claude
Does the AI search the web? When writing context documents, if web research is on for the site (it can be turned off). The searches run at your provider. Claude
How is our API key protected? Encrypted with your company's own key, never shown again after it is entered, used only for your work. Claude

Access

Question Answer
Do you support MFA? Yes: an authenticator app, asked for before the password; your company can require it for everyone. Access
Single sign-on? Planned (Microsoft Entra ID and others), not yet available. Access
Password policy? At least 12 characters, checked against breached passwords, stored with Argon2id, no forced expiry (NIST SP 800-63B). Access
Session timeout? 12 hours without use, 7 days in any case. Access
Is there an audit log? Yes: sign-ins, changes, exports and approvals, hash-chained so tampering shows, exportable. Access
Role-based access? Yes: company admin, site admin, engineer, viewer, checked on every request. Access

Operations

Question Answer
Backups? Continuous (at most 15 minutes at risk), encrypted, off-site, restored as a test every quarter. Operations
Recovery targets? RPO 15 minutes, RTO 4 hours. Operations
Patching? Operating system security updates automatically; the application from its own tested and scanned images. Operations
Vulnerability scanning? Dependencies (pip-audit) and container images (Trivy) on every release. Application
Breach notification? Within 72 hours of our becoming aware. Incidents
Subprocessors? Cloudflare, Postmark, GitHub (code only), Tailscale (administration only), Bluehost (this website only). Providers
Business continuity? Restore from encrypted off-site backups to a new machine; targets as above. Operations