Security & Trust Center
Autonomous doesn't mean uncontrolled.
Every Viralitea workspace is designed around isolation, permissions, auditability, and human oversight.
Security Principles
Five things that hold, whatever the agents decide.
Viralitea is designed to make marketing decisions independently while keeping data, permissions, and execution under defined controls.
Isolated by Design
Customer workspaces operate independently.
Permissioned Autonomy
Agents operate within defined permissions and boundaries.
Auditable Actions
Decisions and executions can be reviewed through logs.
Human Oversight
High-impact actions can require approval.
Data Responsibility
Customer data is handled according to documented policies.
Tenant Isolation
Your workspace. Your data. Your boundaries.
Workspaces are separated so that one customer's account cannot shape another's.
- Workspace-level data separation.
- Access restrictions.
- Independent campaign context.
- Customer-specific memory.
- No cross-customer campaign visibility.
Access Controls
Who can do what, by role.
The role model each workspace is built around. Availability of individual controls is confirmed during onboarding.
| Role | Typical Access |
|---|---|
| Owner | Full workspace and billing control |
| Admin | Campaign and team management |
| Operator | Campaign execution and monitoring |
| Analyst | Reporting and intelligence |
| Viewer | Read-only access |
Audit Trails
Every important action should leave a trail.
- Agent decisions.
- Campaign changes.
- Budget changes.
- Approval requests.
- Human overrides.
- Experiment launches.
- System alerts.
- User activity.
Example audit record
- Action
- Budget Reallocation
- Agent
- Kuber
- Previous Budget
- ₹50,000
- New Budget
- ₹65,000
- Reason
- Improved qualified lead efficiency
- Approval
- Required
- Status
- Approved
- Timestamp
- 12:42 PM
AI Governance
AI should be accountable for what it does.
- Defined agent responsibilities.
- Permission boundaries.
- Action thresholds.
- Confidence scoring.
- Risk scoring.
- Approval workflows.
- Human override.
- Emergency pause controls.
- Decision history.
Human Oversight
Autonomy is a dial, not a switch.
Four levels. A workspace can be moved up or down at any point, and high-impact actions can require approval at every one of them.
- 01
Observe
Agents analyse and recommend.
- 02
Assist
Agents prepare actions for approval.
- 03
Execute
Agents perform approved actions.
- 04
Optimize
Agents continuously improve within defined limits.
Compliance & Policies
The documents, and the ones still being written.
No certification is listed here. Any that Viralitea holds will appear on this page once it has actually been obtained, and not before.
In preparation
- Acceptable Use Policy
- Subprocessor Disclosure
- Responsible Disclosure Policy
Security questions that these do not answer yet can go to the enterprise team, who will answer them in writing.
Questions
The six things evaluators ask first.
These are the uncertainties a dedicated trust page exists to remove, answered in the order they usually arrive.
How is customer data handled?
Storage, transmission, backup, retention, deletion and access logging are documented internally. Specific standards are not published on this page until each has been technically verified, and the security overview covering them is available on request.
Can one customer's data influence another customer's workspace?
Workspaces are separated at the data layer, with independent campaign context, customer-specific memory and no cross-customer campaign visibility.
What are AI agents permitted to do?
Agents operate inside defined responsibilities, permission boundaries and action thresholds, and the autonomy level of a workspace decides whether they recommend, prepare, execute or optimize.
How are actions approved or reversed?
High-impact actions can require approval before they run. Approval requests, human overrides and emergency pause controls are part of the governance model, and every one of them is recorded.
Who is accountable for AI-generated decisions?
Every important action leaves an audit record naming the agent, what changed, the reason, whether approval was required and who granted it.
Can Viralitea meet enterprise security requirements?
Security questionnaires and technical reviews are answered in writing by the enterprise team. No certification is claimed on this page; any that Viralitea holds will be listed here once it has actually been obtained.
Evaluating Viralitea for an enterprise deployment?
Security questionnaires, documentation requests and technical review calls all start in the same place.