Security is the entry ticket.
Foundation runs inside operations where an incident is measured in production, not in headlines. ISO/IEC 27001:2022 certified, independently penetration-tested, built zero-trust from the ground up — and designed so your control system can never be touched by accident. Here are the answers your security team will ask for first.
THE CIO CHECKLIST
The short answers
Where is our data?
In Azure, in a region agreed with you and documented on request. Production data exists only in the production environment.
Who can access it?
Your users, through the application, via your SSO (OAuth2/OpenID Connect, MFA). Named Arundo operations and security staff only, least-privilege, logged.
Is it encrypted?
AES-256 at rest; TLS 1.2+ in transit, end to end.
Can you write to our PLCs?
Not unless you decide to. Data flows one way — out of your site — by default. Write-back exists only at Level 4 of the scale, inside an envelope you define.
Do you train AI on our data?
No. Your data is never used to train models — not ours, not a model provider's. Inference runs inside controlled cloud environments; details per deployment on request.
Are you certified?
ISO/IEC 27001:2022, audited annually by an accredited certification body; certificate and ISMS documentation available on request.
Are you tested?
Independent penetration tests at least annually against web and API surfaces; reports available under NDA.
Who else touches our data?
A short, fixed list of sub-processors — cloud hosting, edge security, identity, code hosting, managed databases, security scanning — provided in full on request. No others.
Zero trust, in five layers
Every layer assumes the others may fail.
- People & process
- background-checked, security-awareness trained and certified, ISMS-documented; secure SDLC with segregated environments and peer-reviewed CI/CD; annual external audit cycle.
- Physical & endpoint
- access-controlled offices; 100% endpoint coverage with real-time analysis and behavioral threat detection.
- Network
- every environment in its own Azure Virtual Network behind Azure Firewall; ingress terminated at a global edge network with a web application firewall (OWASP core ruleset) and DDoS protection; private endpoints for data services.
- Application
- enterprise identity integration with your SSO (OAuth2 / OpenID Connect), MFA, session validation on every request; hardened APIs; container-level RBAC on managed Kubernetes; comprehensive logging.
- Data
- AES-256 at rest and in transit, authentication on every data request, regional hosting, encrypted backups with point-in-time restore — and restore tests, not just backups.
Your data, protected
Three environments (develop, stage, production) with separate access control; live customer data only in production; stage identical to production so what's tested is what runs. Operational data in managed PostgreSQL time-series databases with encryption at rest and in transit, activity monitoring, encrypted backups, periodic restore tests. Secrets and certificates in Azure Key Vault — audited, permissioned, rotated per ISO/IEC 27001:2022 control 8.24. Configuration held as code under version control; every change peer-reviewed and pipeline-deployed.
Your control system, untouched
Data leaves your site one way: pushed from your historian or PLC/SCADA to Foundation's landing zone (Azure IoT Hub, encrypted in transit). No inbound connection to your OT network is required — or, if you decide, allowed. You choose which tags leave; no third party configures or accesses your historian. This is the security architecture of Levels 0–3 of The Six Levels of Autonomous Plant Operations: the system advises, humans act.
Write-back to control exists only at Level 4 — Bounded closed loop, inside an operating envelope agreed with your operations and safety functions, with every permitted action named in advance, every action logged and reversible, and automatic degradation to a safe state outside the envelope.
AI and your data
Your data is never used to train AI models — not by Arundo, not by any model provider. Inference runs inside controlled cloud environments under contractual no-training terms; the specific arrangement can differ per customer and is documented on request.
- AI Crew agents operate on your Domain with the same RBAC as your users — an agent never sees more than the person it works for.
- MCP servers expose the Domain to your own AI clients (Claude, ChatGPT, others) with per-user OAuth 2.1 — every request is authenticated as a named user, authorized per resource, and logged. Your AI client's own data policy governs what happens in that client; Arundo's servers return only what that user may see.
- Documents indexed for document intelligence stay in your instance; embeddings are derived data under the same controls.
Certification and independent testing
ISO/IEC 27001:2022, certified by an accredited certification body; internal audits continuous, independent third-party audits annual. Certificate provided on request. Penetration testing of web application and API surfaces by an independent firm at least annually, OWASP ASVS-based, findings tracked to closure; latest reports available under NDA. GDPR: Arundo acts as processor under a DPA; data residency agreed per customer and documented on request.
Secure development
Version control with protected branches and enforced review; automated dependency-vulnerability and license scanning on every change; infrastructure as code with automated drift detection; build artifacts in Azure Container Registry with Microsoft Entra ID authentication; release branches with locked-down permissions.
Sub-processors and residency
Arundo uses a short, fixed set of sub-processors, each under a data-processing agreement:
- Cloud hosting (Microsoft Azure)
- Edge security
- Identity
- Code hosting & CI/CD
- Managed databases
- Security scanning
The complete named list, and the data residency agreed for your deployment, are provided on request under NDA.
Report a vulnerability
contact@arundo.com — we acknowledge within two business days.