Security built in, not bolted on.⁠

AI can only work for your business if you can trust it. Here is how we protect your data, govern every agent, and keep you in control.

Data ownership

Your data, code, and agents belong to you. Not to us, and not to a model provider.

Your repositories

Source code, agent definitions, prompts, and evaluations are delivered to repositories you control.

Your database

Operational data lives in a database on your infrastructure, fully queryable, with no export fees.

No training on your data

Client data is never used to train external models, by us or by any model provider we use.

Zero exit cost. Any qualified engineer can operate and extend what we build. You can walk away with a working system at any time.

AI governance

Every agent ships with explicit limits, enforced in software rather than in prompts.

  • Scoped permissions

    Each agent has its own service account with access limited to exactly what its job requires.

  • Approval thresholds

    Below a confidence threshold you set, a person decides. Customer-facing actions can require approval indefinitely.

  • Hard limits

    Caps on amounts, pricing floors, and forbidden actions are enforced at the integration layer, where the agent cannot bypass them.

  • Full audit trail

    Every action records what the agent saw, what it decided, which policy applied, who approved it, and which model ran.

  • Reversibility

    Agent actions are designed to be undone, with a clear window and owner for every reversal.

  • Off switch

    Every agent can be paused instantly by your team without disrupting the underlying system of record.

Infrastructure

Deployed where your compliance requirements say it should be.

Encryption

Data is encrypted in transit with TLS and at rest using your provider’s managed encryption.

Your cloud or on-premises

Systems deploy with Docker and Kamal to the cloud provider of your choice, on-premises, or hybrid.

Backups

Automated database backups with defined retention, tested restores, and documented recovery steps. The darker bar is the weekly restore test.

Least privilege

Service accounts, API keys, and deploy credentials are scoped to the minimum access required and stored as secrets, never in code.

Application security

Security controls built into every platform from the first commit.

01Role-based access

Authorization policies (Pundit) control what every role can see and do, down to individual records.

02Audit logs

Change history on sensitive records (Paper Trail) shows who changed what, and when.

03Modern authentication

Passwordless magic-link sign-in, email verification, and secure password reset where passwords are used.

04Abuse protection

Bot protection on public forms and checkout flows, rate limiting, and input validation throughout.

Healthcare & compliance

We build for regulated environments and support the audits that come with them.

HIPAA

Business Associate Agreements are available for any engagement that touches protected health information.

  • PHI stays put

    Agents can run on local models hosted in your environment, so patient data never leaves your control.

  • Audit support

    We support your SOC 2 and HIPAA audits with documentation, evidence, and architecture walkthroughs.

  • Vendor reviews

    We complete vendor security questionnaires and participate in your security review process.

Your choice of model, per workflow

Agents are built on a provider-agnostic layer, so you choose models based on capability, cost, and compliance, and can fail over or switch without changing application code.

  • Anthropic ClaudeCloud
  • OpenAICloud
  • Google GeminiCloud
  • Local modelsVia Ollama, in your environment

Questions security teams ask

Is our data used to train AI models?

No. Client data is never used to train external models. We use model providers under terms that exclude training on customer data, and sensitive workloads can run on local models in your own environment.

Where is our data stored?

In infrastructure you control: your cloud account, your data center, or a hybrid of both. We do not host client data in shared Brem infrastructure unless you ask us to manage it for you.

Will you sign a BAA?

Yes. We sign Business Associate Agreements for engagements involving protected health information, and require the same of any model provider that processes PHI on your behalf.

Do you hold a SOC 2 report?

Brem does not currently publish a SOC 2 report. We design systems to support your own SOC 2 and HIPAA programs, and we complete vendor security questionnaires on request.

What happens if an agent does something wrong?

Every action is logged with its reason and is designed to be reversible. Agents earn autonomy in stages, operate within hard limits enforced in software, and can be paused instantly by your team.

Security questions or a vendor questionnaire?
We respond within one business day.
info@brem.io

Put your agentsto work

Bring your security team to the first call.

We'll walk through architecture, data flows, and agent governance for your environment, and answer your questionnaire.