ISO/IEC 27001
Our Statement of Applicability covers all 93 Annex A controls. Readiness review in 2027.
Certification plannedSecurity and privacy
How we protect your data and your processes
Every measure on this page comes from our controls register and our published terms, and is regenerated when they change. Where something is committed but not yet running, we say so.
Hosted in the EUEU providers only
Processor under the GDPRDPA public
Never trained on your dataBy contract
Frameworks
We build along ISO 27001 and ISO 42001 and plan to certify for both. Until then we show, per control, how we meet them.
ISO/IEC 42001
All 38 Annex A controls mapped, including impact assessment per AI system. Readiness review in 2027.
Certification plannedGDPR
Processor for your data. Public DPA with security measures and retention periods.
In placeNIS2
Measures, provider list, incident notice and audit right for your supply chain duties.
In placeEU AI Act
Human oversight on every change, the model per agent recorded, every action logged.
In placeNEN 7510
A mapping of our measures for your own assessment.
On requestControls
What we do, organised along the Annex A of both standards. Each line shows which controls it covers; hover a number for its name.
In placeIn progressPlanned
A.5Organizational controls37 controls
| Security and AI policy, approved by the directors and reviewed yearly | 5.15.25.4 | In place |
| Every system, service and identity inventoried, checked daily | 5.95.16 | In place |
| Least-privilege access per person, agent and service, reviewed quarterly | 5.155.18 | In progress |
| Every provider checked before use and bound by a GDPR Art. 28 agreement | 5.195.205.22 | In progress |
| Hosted in the EU; Claude moving to AWS Bedrock in the EU | 5.145.23 | In progress |
| Public security advisories matched daily against the software we run | 5.65.7 | In place |
| Incident policy with fixed roles; breach notice within 48 hours, we aim for 24 | 5.55.245.255.265.27 | In place |
| Processor under the GDPR, with a record of processing | 5.315.34 | In place |
| Retention periods per record type, enforced automatically | 5.33 | In progress |
| Recovery within 4 hours and at most 24 hours of data lost, proven by a full restore drill | 5.295.30 | In progress |
| Independent security test every year, and your right to audit us | 5.35 | Planned |
A.6People controls8 controls
| Remote work only over our private network | 6.7 | In place |
| Everyone reports security events to security@joinaitonomy.ai | 6.8 | In place |
| Access removed on a person's last working day | 5.116.5 | In progress |
| Signed confidentiality for everyone who works for us | 6.26.6 | In progress |
| Yearly security and AI awareness session | 6.3 | Planned |
A.7Physical controls14 controls
| No own offices or data centres: physical security sits with our EU hosting providers, which hold ISO 27001 | 7.17.27.47.57.11 | In place |
| Laptops with encrypted disks and screen lock, wiped before re-use | 7.77.97.107.14 | In progress |
A.8Technological controls34 controls
| Sign-in with your own Microsoft or Google account, two-step verification checked | 8.5 | In place |
| Your data separated from other customers in the database, tested on every change | 8.38.27 | In place |
| Every uploaded file scanned for malware | 8.7 | In place |
| Personal data redacted from answers by default | 8.118.12 | In progress |
| Tamper-evident audit log, checked daily | 8.15 | In place |
| Checked from outside every minute, with alerts and a public status page | 8.16 | In place |
| Daily encrypted backups at a separate EU provider, protected for 28 days | 8.13 | In place |
| TLS on all traffic; backups and uploads encrypted | 8.218.24 | In place |
| Every change reviewed and checked automatically; everything as code | 8.98.258.288.32 | In place |
| Dependencies updated automatically, bill of materials for every release | 8.8 | In place |
| Synthetic data only in development and testing | 8.318.33 | In place |
| Deletion of your data confirmed with a certificate | 8.10 | In progress |
| Database and volumes encrypted with EU key management | 8.24 | Planned |
| Admin access only over a private network, no shared or root accounts | 8.28.208.22 | In progress |
A.2Policies related to AI
| AI principles in our policy: a person decides, no training on your data, EU models | A.2.2A.2.3A.2.4 | In place |
A.3Internal organization
| Roles fixed per process: you decide the use, we run the service | A.3.2 | In place |
| Concerns about AI behaviour reported to security@joinaitonomy.ai | A.3.3 | In place |
A.4Resources for AI systems
| Model and hosting recorded per agent | A.4.2A.4.5 | In place |
| Every tool classified as read or write; unclassified tools refused | A.4.4 | In place |
| Data per source and field documented and limited per connection | A.4.3 | In progress |
A.5Assessing impacts of AI systems
| AI impact check per process, alongside your DPIA | A.5.2A.5.3A.5.4A.5.5 | In progress |
A.6AI system life cycle
| Agent definition per process: purpose, tools, rights and approval rules | A.6.2.2A.6.2.3 | In place |
| Evaluation set per agent and safety tests on every change | A.6.1.3A.6.2.4 | In place |
| You approve an agent's autonomy before it acts on its own | A.6.2.5 | In place |
| Every action logged with the model used | A.6.2.6A.6.2.8 | In place |
A.7Data for AI systems
| Never trained on your data | A.7.2 | In place |
| Data only from your own systems, on your grant | A.7.3 | In place |
| The record an agent touched is logged | A.7.5 | In place |
| Retrieved content cleaned and marked as data, not instructions | A.7.6 | In progress |
A.8Information for interested parties
| Plain-language summary, product sheets and public terms | A.8.2A.8.5 | In place |
| Incident notice within 48 hours, we aim for 24 | A.8.4 | In place |
| Anyone can report adverse impacts to security@joinaitonomy.ai | A.8.3 | In place |
A.9Use of AI systems
| Every change approved by a person; no decisions about people on their own | A.9.2 | In place |
| Intended use set per agent; high-risk use agreed in writing first | A.9.3A.9.4 | In place |
A.10Third-party and customer relationships
| Roles fixed per configuration in our DPA and terms | A.10.2A.10.4 | In place |
| Model providers checked before use, from their own documents | A.10.3 | In place |
Documents
Our terms are public. Detailed security documents are shared on request, under a confidentiality agreement where needed.
| Document | What it covers | Access | |
|---|---|---|---|
| Data Processing Agreement | Our role, security measures (Schedule 1), retention periods | Public | Open |
| Sub-processors | Every provider, location and transfer basis | Public | Open |
| SaaS Terms | The service, responsibilities, exit and export | Public | Open |
| Service Level Agreement | Support hours, response times, maintenance | Public | Open |
| Security summary | This page in plain language, as a PDF | On request | |
Email hello@joinaitonomy.ai with the subject Document request: Security summary and your company name. We send it within two business days, under a confidentiality agreement where needed. | |||
| Statement of Applicability | ISO 27001 and ISO 42001 controls, status and evidence | On request | |
Email hello@joinaitonomy.ai with the subject Document request: Statement of Applicability and your company name. We send it within two business days, under a confidentiality agreement where needed. | |||
| Security and incident policies | How we work, and what happens in an incident | On request | |
Email hello@joinaitonomy.ai with the subject Document request: Security and incident policies and your company name. We send it within two business days, under a confidentiality agreement where needed. | |||
| DPIA input pack | Data flows, risks and measures for your DPIA | On request | |
Email hello@joinaitonomy.ai with the subject Document request: DPIA input pack and your company name. We send it within two business days, under a confidentiality agreement where needed. | |||
| NEN 7510 mapping | Our measures against NEN 7510, for healthcare | On request | |
Email hello@joinaitonomy.ai with the subject Document request: NEN 7510 mapping and your company name. We send it within two business days, under a confidentiality agreement where needed. | |||
| Penetration test summary | Findings and fixes of the yearly independent test | After first test | Planned |
Sub-processors
Every provider that processes your data, all in the EU. Changes are announced 30 days ahead.
| Provider | What for | Where |
|---|---|---|
| Scaleway SAS | Hosting, storage, email for notifications, open-weight models | France, Netherlands |
| Hetzner Online GmbH | Encrypted backups and recovery | Germany |
| Amazon Web Services EMEA SARL | Claude via Amazon Bedrock, EU region | EU region |
| Mistral AI SAS | AI models | France |
| TransIP B.V. | Our business email | Netherlands |
| Lettermint B.V. | Standby email for notifications | Netherlands |
Common questions
The questions buyers and security teams ask us most.
Where is our data, and does it stay in the EU?
In the EU: European hosting, backups and AI models. A fully European chain is possible.
Who at Aitonomy can see our data?
Only people who need it for the service, from the EEA, on personal accounts. Working in your environment only with your permission, for a limited time, and logged.
How long do you keep data, and what if we stop?
Fixed periods in our DPA; we keep no copy of your source data. On exit: everything in an open format within 10 business days, no fees, then deletion confirmed in writing.
Is our data used to train AI?
No. Excluded in our contracts and in those of our providers.
How do people sign in?
With your own Microsoft or Google account, so your two-step verification policy applies. Only people you invite get in.
What can an agent do?
Only what you grant, per person and per system. Through Connect never more than the person using it.
Where is the human in the loop?
Every change is a proposal a person approves. Agents do not decide about people on their own.
How do you prevent manipulation of an agent?
Several layers: human approval, screening of content and actions, minimal rights, continuous testing.
Which AI models do you use?
Per task: European models (Mistral and open models) and Claude via AWS Bedrock in the EU. Never trained on your data.
What happens in an incident?
We inform you within 48 hours at the latest and aim for 24, following a written procedure, so you can meet your own deadline.
Are you ISO 27001 or SOC 2 certified?
Not yet. Certification for ISO 27001 and ISO 42001 is planned, with a readiness review in 2027. Until then we build along both standards, have an independent test every year, and you may audit us once a year.
How do you help us with NIS2?
With our measures, provider list, incident notice, audit right and a fully European option where required.
Report a vulnerability
Found a security issue? Email us and we will respond on the next business day. Please do not test against customer environments.