Home · Trust & Security
Trust by design

Audited. Annually renewed.

Your concept notes, consortium maps, and unsubmitted drafts are some of the most sensitive IP your team produces. Every system, process, and model-routing option is designed to make the data path clear before you submit sensitive proposal content — certified to ISO/IEC 27001:2022 by an accredited third party.

ISO/IEC 27001:2022 certified General Data Protection Regulation aligned Encrypted at rest & in transit Model routing controls
Download certificate ↓ Procurement summary ↗ Live compliance dashboard ↗
Issued byGCAI CERTIFICATION SERVICES LLP
Certificate
of Registration
ISO/IEC 27001:2022
Information Security Management System
Awarded to Healthdev OÜ trading as WinGrants AI
Certificate №GCAI-ISMS-LP84
Valid until2029-03-22
First issued2026-03-23
ScopeWinGrants AI ISMS
— Authorized Signatory, GCAI
ISO
27001
2022
What ISO 27001 actually means

A system, not a checkbox.

ISO/IEC 27001 certifies that we operate an Information Security Management System (ISMS) — a documented, audited, continuously improving way of handling information security risk. 4 commitments underpin it.

01

Risk-based

Every control we run traces back to a specific, documented risk. Risks are reviewed quarterly by the CISO and board.

02

Documented

Policies, procedures, records. Auditors read them, staff follow them, exceptions are logged. No verbal traditions.

03

Audited

Internal audits quarterly. External surveillance audits annually. Full recertification every 3 years. By an accredited body.

04

Improving

Every incident, near-miss, and finding feeds a corrective-action register. Controls get stronger each cycle — never weaker.

The 93 controls, grouped

ISO 27001:2022 Annex A, at a glance.

ISO/IEC 27001:2022 organises 93 security controls into 4 themes. Here's what sits under each — and what we do about it.

A.5 · 37 controls

Organisational

Policies, roles, suppliers, incident management, BC/DR.
  • A.5.1ISMS policy. Versioned in Git, reviewed quarterly by the board. Last revision v4.2, March 2026.
  • A.5.7Threat intelligence. MISP feed + NVD + vendor advisories; triaged daily by the SOC.
  • A.5.19–5.23Supplier security. Written DPA + SIG-Lite review on every subprocessor. Annual reassessment.
  • A.5.24–5.28Incident response. Playbook-driven. 4h customer notification SLA. 72h regulator per GDPR Art. 33.
  • A.5.30ICT continuity. RPO 1h · RTO 4h. Cross-AZ replication, quarterly DR tabletop exercises.
  • A.5.33Records protection. Immutable WORM backups 90 days, geographically distributed across EU.
A.6 · 8 controls

People

Screening, employment, awareness, off-boarding.
  • A.6.1Screening. Background & criminal-record checks on every employee and contractor pre-hire (EU BGH-compliant).
  • A.6.2Terms of employment. NDAs + security clauses in every contract; signed before access is provisioned.
  • A.6.3Awareness training. Mandatory annually + onboarding. Quarterly phishing drills (click rate < 3%).
  • A.6.4Disciplinary process. Documented, published to staff, invoked on any policy breach.
  • A.6.5Termination. Access revoked within 24h of off-boarding, hardware wiped & returned.
  • A.6.6Confidentiality. NDAs survive employment; customer project access is audit-logged per-touch.
  • A.6.7Remote working. Zero-trust: MDM-enforced devices only, no split-tunnel, FIDO2 required.
A.7 · 14 controls

Physical

Facilities, equipment, clear-desk, secure disposal.
  • A.7.1–7.4Physical perimeters. Tier III+ EU data centres (OVHcloud RBX, Hetzner FSN). Biometric + badge, CCTV 90-day retention.
  • A.7.5Environmental. N+1 power, fire suppression (Novec 1230), leak detection. SLA 99.982%.
  • A.7.9Off-premises assets. Full-disk LUKS/FileVault, remote wipe via Jamf/Intune, 15-min idle lock.
  • A.7.10Storage media. No removable media. Decommissioned drives crypto-shredded (NIST 800-88 purge) + physically destroyed; certificate of destruction retained.
  • A.7.13Equipment maintenance. Vendor access supervised, data sanitised before RMA.
  • A.7.14Secure disposal. Certificate of destruction issued per-device; logged in CMDB.
A.8 · 34 controls

Technological

Crypto, IAM, logging, secure SDLC, vulnerability mgmt.
  • A.8.2–8.5Identity & access. SSO via SAML 2.0 & OIDC, SCIM 2.0 provisioning, MFA mandatory, FIDO2/passkeys. Break-glass accounts logged + reviewed monthly.
  • A.8.11Data masking. PII pseudonymised in logs & analytics. Per-tenant envelope encryption.
  • A.8.12Data leakage prevention. Egress allow-list on prod, DLP on endpoints, honey-tokens in repos.
  • A.8.15–8.17Logging & monitoring. Signed immutable logs (Grafana Loki + append-only S3). SIEM: Elastic Security. 24/7 SOC. Clock sync via NTS.
  • A.8.24Cryptography. AES-256-GCM at rest, TLS 1.3 (CNSA suites) in transit. HSM-backed KMS (AWS CloudHSM FIPS 140-2 L3 equivalent in EU).
  • A.8.25–8.31Secure SDLC. Mandatory code review, SAST (Semgrep), DAST (ZAP), SCA (Dependabot), SBOM per build (CycloneDX), signed commits & images (Sigstore/cosign).
  • A.8.28Secure coding. OWASP ASVS L2 baseline. Quarterly external pen-tests by an EU-based, BSI-recognised provider. Public bug bounty programme.
  • A.8.32Change management. GitOps + protected branches + peer review; prod deploys require 2-of-N approval.
The audit cycle

How we stay certified.

1
Every quarter

Internal audit

CISO-led audit of a rolling sample of controls. Findings feed the corrective-action register.

2
Every year

Surveillance audit

External accredited auditors spot-check controls on-site and remotely. Report lands within 4 weeks.

3
Every 3 years

Full recertification

Complete re-audit of every applicable Annex A control. A new certificate is issued for the next cycle.

4
Continuously

Monitoring

24/7 SIEM, quarterly external pen-tests, and a public bug bounty programme feed the system between audits.

Why our posture is stronger

Model routing controls. Clear by design, documented by contract.

WinGrants AI separates application hosting, storage, and model inference. Self-Service supports bring-your-own-keys (BYOK). AI Enablement adds tenant-scoped key management. Enterprise can be deployed with approved sovereign inference or fully local campus compute.

Where your draft lives.

Public LLM ✗US-hosted, CLOUD Act jurisdiction
R&I Consultancy ✗Their workspace, their backups, their sub-processors
WinGrants AI ✓European Union-hosted application by default · model route selected by user, tenant, or Enterprise deployment

Training-data exclusion.

Public LLM ✗Opt-out, sometimes — depends on tier and TOS version
R&I Consultancy ✗No guarantee — they use the tools they use
WinGrants AI ✓Customer proposal content is not used to train WinGrants AI. Provider handling depends on selected model route and is documented in the plan or agreement, but all our vendors follow a zero data retention (ZDR) policy.

Audit and verifiability.

Public LLM ✗Closed weights · no audit hook · no signed logs
R&I Consultancy ✗Self-attested — usually nothing in writing
WinGrants AI ✓ISO 27001. Annually audited ISMS · signed immutable logs
The routing layer

Enforced at the routing layer, applied to your research IP.

Our LLM routing layer is built on a zero-retention pattern — no model trains on your content, no unauthorised data egress, and every call is scoped and logged.
Self-Service is EU-hosted & ISO27001 certified;
AI Enablement adds tenant-scoped routing and bring-your-own-keys (BYOK);
Enterprise can pin routes to approved providers or fully local inference.

The same posture covers every Horizon Europe programme — RIA, IA, CSA, EIC, ERC. 

Running a security review? Request a 30-min briefing → — we'll walk through the routing architecture with your IT team.
For your security review

What auditors usually ask first.

Who certified you, and under which accreditation? +

Our certifying body is accredited under the International Accreditation Service (IAS); certificate status is verifiable on iafcertsearch.org. The certificate is mutually recognised across Europe and most of the world.

Which version of ISO 27001 do you hold, and what's in scope? +

ISO/IEC 27001:2022, covering the WinGrants AI platform, our development pipeline, internal corporate IT, EU data-centre footprint, and support operations. The full scope statement and Statement of Applicability are available under NDA.

How does the certification cycle work? +

External surveillance audits annually, full recertification every 3 years, with internal audits in between. Findings flow into a corrective-action register reviewed by management.

How do you handle GDPR? +

We act as a data processor under a signed DPA, EU-hosted by default. ISO 27001 governs the security system; GDPR governs personal-data rights — both apply.

Is our data used to train your models? +

No. Customer proposal content is not used to train WinGrants AI. For third-party model providers, handling depends on the selected route and provider terms. Restricted and Enterprise configurations can be scoped to approved providers or local inference.

Can we run WinGrants AI on our own infrastructure? +

Yes. Enterprise has two deployment paths: Enterprise Lite for on-premise application and data with approved inference, and Enterprise Campus Compute for fully local model inference where sufficient graphics processing unit capacity exists. Reach out at aj@wingrants.ai to discuss scope.

What happens in the event of a security incident? +

We follow a documented incident-response playbook aligned with ISO A.5.24–5.28 and GDPR Art. 33: customer notification through your designated contact, regulator notification where applicable, forensic preservation, and a post-mortem to affected customers. Specific timelines are set in the DPA.

How is data encrypted, and can we bring our own keys? +

Envelope encryption with per-tenant Data Encryption Keys protected by a KEK held in a hardware-backed key store. Configurable model routing is available on AI Enablement and Enterprise. Customers can use OpenRouter-managed commercial routing, their own approved model accounts, approved European endpoints, or fully local inference under Enterprise Campus Compute. Bring Your Own Key (BYOK) is supported when customers want to route usage through their own approved model accounts or endpoints. HYOK (hold-your-own-keys) is reserved for air-gapped Enterprise deployments. Specifics are confirmed under NDA.

How is identity managed? Can we plug into our IdP? +

Modern protocols (SAML 2.0, OIDC, SCIM 2.0) are supported on Enterprise, with JIT provisioning and group-to-role mapping. MFA is mandatory, FIDO2/passkeys supported. Specific IdP integrations are confirmed during scoping.

How do we get our data back — or delete it — at contract end? +

Self-serve export in JSON and DOCX, or an admin-initiated full export and delete. Deletion timelines (including backups) and any certificate of erasure are documented in the DPA.

How can we get the full audit report and SoA? +

Statement of Applicability, audit reports, and pen-test summaries are available under a short mutual NDA. Email aj@wingrants.ai.

Prefer to walk through this in person? For RTOs, universities and EU agencies, we travel to your site to brief your IT and security team directly. No travel charge for EU institutions.
Arrange a briefing →