Skip to content

Security and trust

Powerful automation needs firm boundaries

Social Agent is designed so that AI and workflows can act on real systems without putting them at risk. Every action is permitted, approved where it matters and recorded.

Start free Talk to sales
Workflow activity: every run with status, trigger, start time, duration and steps

Security and trust

Safety by design

  • Secrets never exposed

    Credentials are encrypted at rest, injected at run time and never shown to the AI, the browser or logs.

  • Role permissions

    Owners, admins and members, with per-integration permissions on top.

  • Human approvals

    Require sign-off for sends, posts, database writes and server commands.

  • Audit trail

    Who did what, when, through which integration, including every AI tool call.

  • Consent and opt-outs

    Contact consent and opt-out state enforced on every broadcast.

  • Tenant isolation

    Every query is scoped to your workspace. Data never crosses between customers.

Approvals

Nothing important happens without a yes

Approval steps pause a workflow or an agent action until the right person approves. Approvers see exactly what will be sent, posted, written or run.

  • Approvals for AI drafts and workflow steps
  • Named approvers and expiry times
  • Clear preview of the action
  • Decisions recorded with the reason
Social Agent dashboard with WhatsApp, social media, workflow and email summaries and items that need attention

Audit

A record you can hand to an auditor

Every run, step, message and approval is logged with the person or workflow responsible. Use it to answer "who sent this?" in seconds.

  • Run history for every workflow
  • Per-step inputs and outputs, with secrets masked
  • Audit log of settings and permission changes
  • Run details linked from the calendar and dashboard
Workflow activity: every run with status, trigger, start time, duration and steps

POPIA

Built for South African privacy law

Consent state, opt-outs and erasure requests are first-class features, which helps you meet POPIA obligations for direct marketing and data subject requests.

  • Opt-in and opt-out state per contact
  • Automatic opt-out keyword handling
  • Erase a contact and anonymise their messages
  • Opt-outs respected by broadcasts and automations alike

One platform

Works with the rest of Social Agent

Every product shares the same integrations, permissions, approvals and audit trail.

Guides

Learn more on the blog

FAQ

Security and trust: common questions

Still unsure? Ask our team.

Is Social Agent POPIA compliant?

Social Agent provides the controls you need, such as consent records, opt-out handling and erasure, and acts as an operator on your behalf. Compliance also depends on how you collect consent and use the platform, so speak to your information officer about your own obligations.

Can the AI access my credentials?

No. The AI never receives passwords, tokens or keys. Tools run on the server with the stored credential and only the result is returned to the model.

Who can see what in a workspace?

Roles decide what each person can manage, and integrations can be limited to specific people. The audit log shows every change to these settings.

How do I report a security issue?

Email [email protected] with the details. We take every report seriously and respond as quickly as we can.

Automate with confidence

See how approvals, permissions and audit work in your own workspace.

Start free Book a demo