Security & Trust
DataFuseAI helps teams build and run data pipelines across managed cloud, private-hosted, and fully on-premise environments. This page explains how access control, audit logging, credential handling, isolation, and deployment security actually work today, section by section.
Built Around Access Control, Not an Afterthought
DataFuseAI connects to business-critical databases, warehouses, APIs, and files, so who can see what, who can change what, and where the platform physically runs are treated as core product decisions rather than a checklist added later.
Group-based, action-level RBAC
Permissions are ticked module by module — Pipeline, Cluster, Driver, Query Editor, and more — and assigned to a group rather than to a person directly, so what a user may run, edit, or view is deliberate, not implicit.
Credentials stored with restricted access
Database credentials, API keys, OAuth tokens, SSH keys, and certificates are stored using encryption and access-restriction controls, with least-privilege and rotation recommended for every credential you connect.
Audit logs with before/after detail
Every logged action carries who, when, what changed, and from what IP — including old and new values on the Changes tab — filterable by category and exportable.
Isolated tenants and user workspaces
Each tenant is its own workspace with its own users and groups, and every user gets a private folder with no access to anyone else's space by default.
Encrypted in transit, encrypted at rest
Connections use TLS in transit, and hosted-service systems encrypt data at rest. For private-hosted and on-premise deployments, at-rest encryption follows your own infrastructure controls.
Deploy where your data has to live
Run DataFuseAI fully managed, private-hosted on your own infrastructure, or entirely on-premise with no internet dependency — see deployment options.
Choose Where DataFuseAI Actually Runs
Role-based access control, tenant isolation, and secure credential handling apply regardless of hosting model. Execution engine — the DataFuseAI native engine, a Databricks workspace, or a Spark cluster reached through Apache Livy — is a separate choice from where the platform itself is deployed. In every model, data stays in your own databases, warehouses, and storage; DataFuseAI connects through drivers and profiles rather than relocating it.
Managed
We host and operate DataFuseAI for you — application servers, upgrades, patches, and ongoing maintenance are handled without exposing your data. Fastest path to running your first pipeline.
Private-hosted
Install DataFuseAI on your own private servers or cloud environment. We assist with initial setup; your team owns day-to-day operations, networking, and infrastructure security from there.
Local on-premise (offline)
Run DataFuseAI entirely inside your internal network, with no external hosting and no internet dependency — for environments where data cannot leave the premises and security policy allows no exceptions.
Compliance frameworks bind your organization, not a piece of software — no deployment model produces compliance by itself. See the full comparison on Deployment Options and the underlying platform architecture.
Encryption, Credentials, and Retention
The specifics below vary by deployment model — a hosted-service default is not the same guarantee as a control your own infrastructure team enforces on a private-hosted or on-premise install.
Connections between the browser, DataFuseAI services, and connected databases or APIs use TLS in transit. On the managed hosted service, systems encrypt data at rest. For private-hosted and on-premise deployments, at-rest encryption is provided by your own infrastructure and storage configuration, since that environment is under your control rather than ours.
Database credentials, API keys, OAuth tokens, SSH keys, and certificates are stored using security controls such as encryption and access restriction. We recommend customers follow least-privilege and regular rotation practices for the credentials they connect.
On the hosted service, defaults are: account data for 14 days after termination, audit logs for up to 1 month, system/application logs for 14 days, query history and pipeline/job run history for 7 days each, and soft-deleted items for 7 days before permanent removal. For private-hosted and on-premise deployments, retention is primarily controlled by your own environment. Full detail lives in the Privacy Policy.
On the hosted service, backups are retained for up to 7 days, and deleted data may remain in a backup until that rotation completes. For private-hosted and on-premise deployments, backup and disaster-recovery responsibility is shared with — or fully owned by — the customer, depending on the deployment agreement; we recommend discussing recovery objectives during deployment planning.
Who Can Do What, and a Record of It
Access control and audit logging are the same layer across every deployment model — they don't change depending on whether DataFuseAI is managed, private-hosted, or on-premise.
Permissions are action-specific and grouped by module — Pipeline, Cluster, Driver, Logger, Dashboard, Query Editor, Connection Profile, and more — with a Select All per module. Users don't hold permissions directly; they're assigned to a named group, and a second, file-level layer (Read, Write, Manage, Remove Access) controls which specific files or resources a group can act on. See the full model on platform architecture.
Each entry records When, Who, Category, Action, Resource, Changes, Description, and IP Address. The Changes detail includes old and new values field-by-field — populated even on a create event — so a change can be reviewed, not just noticed. Logs are filterable by category and exportable, and are covered further on Data Governance & Compliance.
Job and pipeline run history is kept alongside execution status, so a run can be traced back to when it happened and what configuration produced it, instead of relying on someone remembering to check a script output.
Isolation, Network, and Shared Responsibility
Isolation is enforced at the platform layer in every deployment model; where the infrastructure boundary itself sits — and who operates it — depends on whether you're on the managed service, private-hosted, or on-premise.
Each tenant operates as its own isolated workspace with its own users, groups, and configuration, managed at the company level by a SuperAdmin role. Beyond that, every user is given a private folder the moment their account exists, scoped by filepath and covering their own connection profiles, jobs, pipelines, and saved queries — the default is no access to anyone else's space.
Your data stays in your own databases, warehouses, and storage — DataFuseAI connects through drivers and connection profiles rather than relocating it. For private-hosted and on-premise deployments, firewalls, VPC configuration, and inbound/outbound network rules are controlled by your own infrastructure team; for the managed service, network security is handled as part of the hosted environment.
Security is shared. We're responsible for the platform components under our control — access control, audit logging, tenant isolation, and credential storage. Customers remain responsible for securing their own source systems, user access decisions, exports, and downstream storage — and, on private-hosted or on-premise deployments, for the underlying infrastructure and network itself.
Development Practices and Incident Response
Security work doesn't stop at launch. Here's how changes to the platform are made, how issues are weighed, and what happens if something goes wrong.
Secure software development
Changes to the platform go through source control, code review, and staged release rather than shipping directly to production — the same discipline that underpins the access-control and audit-log layer customers rely on.
Vulnerability management
Reported issues are evaluated by severity, exploitability, and potential customer impact, and fixes are prioritized accordingly. If you believe you've found a security issue, please report it privately rather than disclosing it publicly before we've had a chance to investigate.
Security incident response
If an incident materially affects customer systems or data, affected customers are contacted in line with applicable contractual and legal requirements. Report a suspected issue to security@datafuseai.com.
Built to Support Your Compliance Program
Compliance frameworks bind your organization, not a piece of software. What DataFuseAI provides is the access-control, audit-log, and isolation layer your team draws on when building toward frameworks like SOC 2, ISO 27001, HIPAA, or GDPR — see how that plays out on Data Governance & Compliance.
Product security controls
Implemented in the platform today
- Group-based, action-level role-based access control
- Audit logging with before/after change detail, filterable and exportable
- Isolated tenant workspaces and per-user private folders
- TLS encryption in transit; encryption at rest on the hosted service
- Credentials and secrets stored with encryption and access restriction
- Managed, private-hosted, and fully offline on-premise deployment options
Working with your compliance team
Documentation, questionnaires, and reviews
If your evaluation needs a completed security questionnaire, a specific compliance framework discussed, or a deeper architecture review, our team can work through it directly with you and your security or procurement team.
Subprocessors: DataFuseAI maintains a list of subprocessors and will provide it upon request; see the Privacy Policy for the current process. Data residency: depending on your subscription and deployment model, you may be able to select a hosting region for the managed service, and private-hosted or on-premise deployments give you full control over where data physically sits.
Have a Security or Deployment Question?
Enterprise evaluations often need a deeper look at deployment architecture, network requirements, credential handling, and data flow than a single page can cover. We're happy to walk through your environment before you commit.
Enterprise security review
Talk to us about deployment architecture, network requirements, RBAC configuration, or data residency for your evaluation — before or during a proof of concept.
Report a security issue
If you believe you've found a security vulnerability, contact us privately. Please include a description of the issue, the affected component, reproduction steps where possible, and potential impact — and avoid accessing, modifying, or exposing data belonging to other users while investigating.
security@datafuseai.com