Legal
Data Policy
Effective date: 8 September 2026 · Last updated: 8 September 2026
Our clients hand us production data, internal documents and system access so we can build AI into their operations. This policy sets out how that data is classified, stored, processed, retained and deleted, and what we will and will not do with it.
01Purpose and scope
This policy applies to all data that Ionio holds or processes, including data belonging to clients, prospects, employees, contractors and website visitors. It covers every environment we use to do the work: laptops and endpoints, cloud infrastructure, code repositories, collaboration tools, and the AI platforms and APIs used in delivery.
It complements our
Privacy Policy, which describes how we handle personal information as a controller, and our
Terms and Conditions. Where a client's own data processing agreement or security schedule imposes stricter requirements, those requirements apply to that engagement.
02Our role: controller and processor
Ionio acts as a controller for data it collects for its own purposes, such as website analytics, sales enquiries, marketing contacts and employment records.
Ionio acts as a processor for data belonging to a client that we handle to deliver the Services. In that role we process only on documented instructions from the client, do not use the data for our own purposes, and return or delete it at the end of the engagement. Where required, we enter into a data processing agreement recording these obligations, the categories of data, the purposes of processing and the approved subprocessors.
03Data classification
Every dataset we handle is classified, and the classification determines where it may be stored, who may access it, and how long it is kept.
Public
Published research, marketing material, public documentation. No access restrictions.
Internal
Operational documents, internal templates, non-sensitive project notes. Available to Ionio personnel.
Confidential
Client business data, source code, architecture, commercial terms, roadmaps. Restricted to the assigned engagement team.
Restricted
Personal data, credentials, financial and health-adjacent records, and anything a client designates as restricted. Named-individual access, additional controls, and logged use.
04Collection and minimisation
- Minimum necessary — we request only the data required for the agreed scope, and we prefer a representative sample over a full production export wherever it will do the job.
- Anonymisation first — where the work does not depend on identifying individuals, we ask for anonymised, pseudonymised or synthetic data.
- Purpose limitation — data supplied for one engagement is not reused for another client, for internal product development, or for benchmarking, without explicit written permission.
- Secure transfer — data is transferred through encrypted channels such as client-managed cloud storage, SFTP or a shared secure workspace. We do not accept credentials or restricted data over unencrypted email or consumer messaging apps.
05Storage and infrastructure
- Client-hosted by default — wherever the client's environment allows it, we build and run inside the client's own cloud accounts so that data never leaves their control boundary.
- Segregation — where we host, each client has logically separated storage, project workspaces, repositories and credentials. Data from different clients is never combined in a shared store.
- Managed providers — infrastructure runs on established cloud providers with recognised security certifications. We do not run production client workloads on personal devices or unmanaged servers.
- Local copies — engineers avoid local copies of restricted data. Where a local copy is unavoidable, it sits on an encrypted disk and is deleted as soon as the task is complete.
- Residency — where a client requires a specific data residency region, we configure infrastructure to keep data in that region and record the requirement in the engagement documents.
06Client data and AI systems
Because we build AI systems, this section matters more than any other. Our standing commitments are:
- We do not train on client data. Client data is not used to train, fine-tune or improve any model for our own benefit or any other client's benefit. Fine-tuning happens only when a client explicitly commissions it, and the resulting model artefacts belong to that engagement.
- Enterprise model terms. We use model and API providers under enterprise or business terms that exclude the data we send from provider training, and we prefer providers offering zero-retention or short-retention processing.
- Consumer AI tools are off-limits. Restricted or confidential client data is not pasted into personal or consumer-tier AI accounts, browser extensions or unapproved tools.
- Prompt and log hygiene. Prompts, completions and traces can contain personal data. Logging is scoped to what is needed to debug and evaluate, sensitive fields are redacted or hashed where practical, and log retention follows the schedule in section 10.
- Retrieval and vector stores. Embeddings and vector indexes derived from client data are treated with the same classification as the source data, stored in the client's environment or a segregated store, and deleted with it.
- Human review. Where a person needs to review AI output containing client data, that review is limited to the assigned engagement team and is subject to the same access controls and confidentiality obligations.
Provider changes. Model providers update their terms and retention behaviour from time to time. We review the terms of the providers we rely on and will inform affected clients where a change materially alters how their data is handled.
07Access control
- Least privilege — access is granted per engagement, per system and per role, limited to what a person needs to do their job.
- Named accounts — no shared logins. Every access is attributable to an individual.
- Multi-factor authentication — required on email, cloud consoles, code repositories, and any system holding confidential or restricted data.
- Client-issued access — where a client provisions accounts in their own systems, we use them rather than creating parallel access paths, and we ask that access be scoped to the minimum required.
- Review and revocation — access is reviewed periodically and revoked at engagement close or on the day a person's involvement ends, whichever is earlier.
08Encryption and secrets
- In transit — TLS 1.2 or higher for all data movement between systems, endpoints and third-party APIs.
- At rest — provider-managed encryption on all storage holding confidential or restricted data, and full-disk encryption on every work device.
- Secrets — API keys, tokens and credentials are held in a managed secret store or the platform's own secret manager. They are never committed to source control, pasted into tickets or documents, or shared over chat.
- Rotation — credentials are rotated on a defined schedule, on personnel change, and immediately on any suspected exposure.
09Subprocessors and vendors
We use a limited set of third-party providers to deliver our services. Categories include cloud infrastructure and hosting, source control and CI, AI model and API providers, observability and error tracking, communication and collaboration, project management, and payment and accounting.
- Every vendor with access to confidential or restricted data is assessed for security posture, data handling terms, retention behaviour and location before use.
- Each is bound by written terms covering confidentiality, security and, where personal data is involved, data protection obligations equivalent to our own.
- A current list of subprocessors used on a given engagement is available to that client on request, and material additions are notified where the engagement requires it.
- We do not sell, rent or trade client data, and we do not share it with any party outside the delivery chain except where compelled by law.
10Retention and deletion
Data is kept only as long as it is needed. Default schedules, unless the engagement agreement sets different terms:
Client production data
Duration of the engagement, then deleted within 30 days of close or on written request.
Working copies and samples
Deleted as soon as the task requiring them is complete.
Application and prompt logs
Typically 30 to 90 days, unless a longer period is agreed for evaluation or compliance.
Backups
Rolling window, after which deleted data ages out of the backup set.
Project documentation
Retained for the support and warranty period, then archived or deleted per the agreement.
Contracts and invoices
Up to 7 years, to meet tax, accounting and legal obligations.
Deletion covers primary storage, derived artefacts such as embeddings and caches, and copies held with subprocessors where we control them. Written confirmation of deletion is provided on request. Data held under a legal hold is retained until the hold is lifted.
11International transfers
Our team and our providers operate across multiple countries, including India, the United States, the United Kingdom and the European Union. Where personal data leaves the European Economic Area or the United Kingdom, we rely on an adequacy decision or on Standard Contractual Clauses with supplementary measures. Where a client requires processing to remain in a specific jurisdiction, that constraint is agreed before work begins and configured into the infrastructure and the choice of providers.
12Security practices
- Secure development — code review before merge, dependency and secret scanning, separation of development, staging and production environments, and no production data in development environments unless anonymised.
- Endpoint security — disk encryption, screen lock, automatic OS and browser updates, and endpoint protection on all work devices.
- Monitoring — logging and alerting on authentication, infrastructure changes and anomalous activity in the environments we operate.
- Vulnerability management — dependencies are monitored and patched on a risk-based schedule, with critical issues prioritised.
- Change management — production changes are reviewed, tested and reversible, with rollback paths defined before release.
- Training — everyone at Ionio receives security and data handling guidance on joining and refreshers as practices change.
13Incidents and breach notification
We maintain an incident response process covering detection, containment, assessment, notification, remediation and post-incident review.
- Containment first — on detection we isolate affected systems, revoke and rotate credentials, and preserve evidence for investigation.
- Client notification — affected clients are notified without undue delay and, where the incident involves their personal data, within 48 hours of confirming it, so that they can meet their own regulatory deadlines.
- What we tell you — the nature of the incident, the categories and approximate volume of data involved, the likely consequences, the measures taken, and a named contact for follow-up.
- Regulators and individuals — where Ionio is the controller, we notify the relevant supervisory authority and affected individuals within the timeframes the law requires. Where we are a processor, we support the client in making those notifications.
- Review — every incident closes with a root cause analysis and corrective actions.
Reporting a vulnerability. If you believe you have found a security issue in a system we operate, report it to us using the contact details in section 19. We will acknowledge the report and will not pursue action against good-faith research that avoids privacy violations, data destruction and service disruption.
14Data subject and client requests
Where we hold personal data as a controller, individuals can exercise their rights as described in our
Privacy Policy.
Where we hold personal data as a processor, requests from individuals are referred to the client without undue delay, and we assist the client in responding, including by locating, extracting, correcting or deleting the relevant records. We do not respond to such requests directly unless the client instructs us to.
Clients can request an export of their data in a standard machine-readable format at any point during an engagement and for 30 days after it closes.
15Personnel and contractors
- Everyone working on client engagements, employee or contractor, is bound by written confidentiality obligations that survive the end of their involvement.
- Access is granted per engagement rather than by default, and is removed on the day involvement ends.
- Devices used for client work must meet our endpoint requirements, and client data is removed from them at engagement close.
- Where a client requires background checks, specific residency, or named-personnel approval, we accommodate that in the engagement terms.
16Backups and continuity
Systems we operate are backed up on a schedule appropriate to the workload, with backups encrypted and access-controlled to the same standard as primary storage. Restore procedures are documented and tested. Where a client's own environment hosts the system, backup and continuity remain under the client's control and their policies apply, with our recommendations documented at handover.
17Audits and assurance
We respond to client security questionnaires and due diligence reviews as part of onboarding. Clients with a contractual right to audit may request evidence of the controls described in this policy once per year, on reasonable written notice, at a mutually agreed time and scope, and subject to confidentiality. Where a control is delivered by a subprocessor, we provide that provider's certifications or reports where they are available to us.
18Governance and review
This policy is owned by Ionio's leadership and reviewed at least annually, and additionally whenever we adopt a materially different tool, provider or delivery model, or when applicable law changes. Revisions are published on this page with an updated effective date. Where a change materially affects an active engagement, the affected client is informed directly.
19Contact
For questions about this policy, a data processing agreement, a security review, or a deletion or export request, contact us using the details below or through the contact form on our website.