Security, privacy and governance built for research teams.

01 · Access controls

Who can get in, and what they can do.

Authentication and access run on managed identity infrastructure. Enterprise identity options are on the roadmap and are labelled as such below.

• Live

  Email and password sign-in

  Sessions are scoped to the user and can be revoked from account settings.

• Roadmap

  Enterprise identity: SSO, SAML/SCIM and MFA

  Planned for Enterprise as one workstream: single sign-on (Google Workspace, Microsoft Entra, Okta), SAML and SCIM provisioning, and TOTP multi-factor authentication. None of these are wired today; sign-in is email and password. Talk to us if any of them is a procurement requirement.

• Live

  Project-scoped access

  Every project is isolated at the database level, with row-level policies enforced by the database itself rather than by application code.

• Roadmap

  Role-based access (owner, analyst, viewer, client)

  A first-class role model with granular permissions is rolling out. Current access is scoped per project and per invited user.

• Live

  Magic-link client portals

  Share findings with clients via a signed link, with no account required on their side. A revocation UI and server-side invalidation of shared links are rolling out alongside the audit log.

02 · Auditability & traceability

Every number should be traceable to its source.

Traceability from an AI-generated statement back to the underlying data is the core trust promise of the platform. A broader audit log is the next milestone.

• Live

  AI outputs link back to source

  Numbers, charts and narratives link to the source table, the exact computation that produced them, the weights applied and the verbatims behind them.

• Roadmap

  Immutable audit log of analyses, exports, shares and portal access

  In development, targeted for Enterprise. Append-only log with actor, project, timestamp and event payload.

• Roadmap

  Audit-log export

  Planned. CSV / JSON export gated to project owners, once the audit log ships.

• Live

  Model-vs-human attribution on new coding runs

  For open-end coding, we record whether a code was assigned by the AI, a human, or a human overriding the AI.

03 · Data handling

Where data lives, and what happens to it.

Data is stored on managed cloud infrastructure. Encryption in transit and at rest is a platform default. Regional options are documented below with realistic caveats.

• Live

  Encryption in transit and at rest

  TLS 1.2+ in transit and AES-256 at rest, provided by the underlying platform.

• Live

  EU and North America hosting

  Choose EU (Frankfurt) or North America (US or Canada) at the workspace level. Existing workspaces can be migrated by request.

• Roadmap

  APAC hosting

  Available on request for Enterprise. We spin up a dedicated project in the target region rather than moving a shared tenant.

• Roadmap

  True regional data residency

  Available via dedicated deployments on Enterprise. It is not offered as a per-project toggle inside a shared tenant.

• Live

  Tenant isolation via RLS

  Every user-facing table is scoped by project ID and enforced with row-level security policies.

• Live

  Your data is never used to train any model

  Your survey data, verbatims and reports are never used to train or fine-tune a model, ours or anyone else's. Training opt-out is set by contract with every AI service involved in processing.

• Roadmap

  Configurable retention and project-close deletion

  A formal retention and hard-delete flow is on the roadmap. Customer data can be deleted on request today.

04 · AI governance

The AI never writes the maths.

Statistics run on a versioned, test-covered engine rather than on a language model. The AI chooses which tested computation to run; a researcher stays in control of what that produces.

• Live

  Statistics run in a versioned statistical engine

  Significance tests, weighting, cross-tabulation and matrix analysis are executed by fixed, version-controlled code covered by a test suite, not generated by a language model. Every result is tied to the specific computation that produced it.

• Live

  Nothing is published without a researcher in the loop

  Every AI-generated table, chart and narrative is a researcher's to review, edit or re-run before it is published or exported. Nothing auto-publishes today.

• Roadmap

  Hallucination QA layer for coding

  In development. Rule-based checks flag themes that don't sufficiently cover source verbatims for reviewer attention.

05 · Governance workflows

Consistency across teams, brands and waves.

The workflow layer is designed so trackers, agencies and multi-brand teams can enforce their standards without depending on a specific analyst remembering to do so.

• Live

  House style applied across every surface

  V²AID binds your brand kit to the deck, the portal and the exports from one profile, so output does not drift between surfaces. A pre-publish check that flags off-brand copy and layout is on the roadmap.

• Roadmap

  Approval gate before external share

  Rolling out. A draft / in-review / approved status will block portal shares until an approver signs off.

• Roadmap

  Wave-on-wave method consistency for trackers

  Rolling out. Analysis plans, weighting and brand kit will be locked at the tracker level so subsequent waves reproduce the same standard.

• Roadmap

  Analysis-plan change history

  Planned. Snapshot and diff view of analysis plans over time, so governance teams can see what changed and why.

06 · Compliance posture

Where we are, honestly.

DataImpel is not yet certified against SOC 2 or ISO 27001. This section describes the current posture without overclaiming. Please ask if you need something specific for procurement.

• Roadmap

  SOC 2 Type II

  Controls documented against the SOC 2 Type II framework; a formal audit is planned. We are not certified today.

• Live

  GDPR-aligned data handling

  EU hosting, tenant isolation, data-deletion on request and no training on customer data. This is not a certification.

• Roadmap

  DPA available on request

  A standard Data Processing Addendum is available on request. Custom terms can be negotiated at Enterprise.

• Live

  Subprocessors list on request

  The current list of subprocessors (hosting, AI services, email) is available on request.

• Live

  Responsible disclosure

  Security researchers can report issues to security@dataimpel.ai. We aim to acknowledge within two business days.