Sign in Get a demo Start free
Security

What we can prove, and what we cannot. Both lists.

Most security pages are a wall of badges. We are a young platform and we do not hold most of those certifications yet, so here is the architecture we can actually point at — and an explicit list of what we do not have, so you can stop reading early if it is a blocker.

Talk to us about your requirements Governance controls
  • Schema-per-tenant isolation
  • Approval gates
  • Full audit trail
Where we stand today
10
Controls in place
5
Named as missing
1
Schema per tenant
0
Badges we do not hold

Both lists, side by side

the second one is the useful one

A vendor who will not tell you what they lack is telling you something.

If a control on the right is a hard requirement for your procurement team, you will find that out in the next thirty seconds rather than in week six of an evaluation.

In place today

Verifiable in the product

  • Schema-level tenant isolation. Every organisation gets its own PostgreSQL schema. Separation is enforced by the database, not by a WHERE clause somebody has to remember.
  • HSTS with preload and subdomains. Enforced TLS, HSTS preload, and an SSL redirect on every request. There is no plaintext path into the application.
  • Hardened session and CSRF cookies. Secure, HttpOnly and SameSite are set on session and CSRF cookies, with content-type sniffing disabled and a restrictive referrer policy.
  • Clickjacking and XSS protections. X-Frame-Options and browser XSS protections are configured at the framework level rather than left to defaults.
  • Role-based access control. Permissions are enforced per organisation, with API-key authentication scoped to the tenant that issued the key.
  • Approval gates on every agent action. Agents draft; humans approve. Nothing is published, sent or spent without passing a gate you control.
  • Full audit trail. Who asked, what was planned, which agent executed, what changed and who approved it — recorded and exportable.
  • Kill switch. One control halts every agent across the organisation immediately.
  • GDPR tooling and backups. Subject-access and erasure handling, plus a managed backup process for tenant data.
  • SOC 2 Type I. Completed. Type I attests that our controls are designed correctly, assessed at a point in time by an independent auditor.

Not yet

Say so before you evaluate

  • SOC 2 Type II. Not held. Type I attests that controls are designed correctly; Type II proves they operated effectively across an observation window of several months. We have not completed that window. Treat any vendor quoting bare "SOC 2" without the type as quoting Type I.
  • ISO 27001 / ISO 42001. Not held. ISO 42001 for AI management systems is on the roadmap because it is the one that will actually matter for AI procurement.
  • Third-party penetration test report. Not yet published. When there is one, it will be available under NDA rather than summarised into a marketing bullet.
  • Guaranteed regional data residency. Not offered as a contractual commitment today. Ask us where your data would physically sit and we will tell you plainly.
  • Customer-managed encryption keys. Not available. Encryption is managed by us, at rest and in transit.

“If a control on the right-hand list is a hard requirement for you, we are not the right platform this year. We would rather you knew now.”

Why both lists are published

How tenant isolation actually works

the part worth understanding

Separation enforced by the database, not by application code.

Most multi-tenant SaaS keeps every customer in shared tables and filters by an organisation column. One missing filter in one query is a cross-tenant data leak. We use a schema per tenant, so the isolation holds even when application code is wrong.

One schema per organisation

Your data lives in its own PostgreSQL schema. A query executed in your context cannot see another tenant's tables at all.

Keys scoped to the tenant

API keys authenticate into a single organisation. A leaked key exposes one tenant, never the platform.

Failure mode is an error, not a leak

When tenant context is missing, the query fails rather than silently returning somebody else's rows.

The AI questions buyers actually ask

newer than most security pages

"What does your AI do with our data?" deserves a direct answer.

Traditional security review covers encryption and access control. It does not cover an agent that can send email on your behalf. These are the controls that address that.

Agents cannot act unilaterally

Every outbound action — sends, publishes, spend changes — queues for human approval by default. Autonomy is granted per agent and per action, never globally by accident.

Everything an agent does is logged

The audit trail covers the request, the plan, the agent that executed, the data it changed and the human who approved. Exportable for your own review.

One kill switch

If something is behaving unexpectedly, one control stops all agent activity organisation-wide. No uninstall, no support ticket, no waiting.

Model providers do not train on your data

Prompts are processed under enterprise API terms that prohibit training on customer content. We will share the specific terms under NDA.

Security questions

answered directly
Is Marketing Titan SOC 2 certified?

No. We do not hold SOC 2 Type II today and we are not going to imply otherwise. If SOC 2 is a hard procurement requirement for you this year, we are not the right platform yet.

How is my data separated from other customers?

Each organisation gets its own PostgreSQL schema rather than sharing tables filtered by an organisation column. Isolation is enforced by the database, so it holds even if application code has a bug. When tenant context is missing a query fails rather than returning another tenant’s rows.

Can an AI agent send an email or spend money without approval?

Not by default. Every outbound action queues for human approval. Autonomy can be raised per agent and per action, and a single kill switch halts all agent activity organisation-wide.

Is my data used to train AI models?

Prompts are processed under enterprise API terms with model providers that prohibit training on customer content. We will share the specific contractual terms under NDA.

Do you offer data residency guarantees?

Not as a contractual commitment today. We will tell you plainly where your data sits, but we will not promise regional pinning we cannot yet enforce.

Can I get a penetration test report?

We do not have a published third-party report yet. When we do it will be available under NDA rather than reduced to a marketing bullet.

Security review

Send us your security questionnaire.

We will complete it honestly, including the questions where the answer is no. If a gap is a blocker we will tell you rather than run you through an evaluation.

Talk to us Governance controls