Skip to content

Security

Security for the most confidential information.

Herfindahl is built for high-stakes work where confidentiality and defensibility are non-negotiable. The product is designed to keep data where it belongs, enforce permissions strictly, and make actions traceable, reviewable, and auditable.

Secure by default. Built in.

Security isn’t an add-on in Herfindahl—it’s built into the product from the ground up. We enforce four principles strictly: keep data where it belongs, enforce permissions rigorously, make actions reviewable, and let customers choose the deployment model that fits their environment.

Trusted data storage

Local-first workspaces

Case files stay in your workspace; Herfindahl doesn’t collect your project into the cloud by default.

Clear boundaries

The product operates within explicitly selected workspaces to reduce accidental cross-case leakage.

Flexible inference routing

Route inference through infrastructure operated by your organization, or use a secure Herfindahl-managed endpoint.

Configurable logging

Supports policies that avoid storing prompts and outputs, keeping logs focused on operational metadata.

No training on case content

Herfindahl does not use your files, prompts, or outputs to train models. Customer-controlled deployment models keep content private and segregated under your organization’s control.

Controls that map to real risk

Least-privilege by design

The assistant can only use the capabilities you expose in your environment.

Human-in-the-loop execution

Sensitive actions require explicit confirmation before they execute.

Encryption & transport security

Protect data in transit and apply encryption at rest where data is stored.

Enterprise authentication

Supports integration with enterprise SSO and modern authentication approaches for controlled environments.

Enterprise assurance

We provide comprehensive security documentation and share security certifications and third-party penetration testing reports with enterprise customers.

Your data. Your decisions.

You maintain control over your data at every step.

Data retention

Case content stays in your workspace, so retention follows your organization’s policies. Logging can be configured to avoid storing prompts and outputs.

Data governance

Explicit workspace scoping and strict permissions help keep work separated by case, with actions designed to be traceable, reviewable, and auditable.

Encryption management

Protect data in transit and apply encryption at rest where data is stored, aligned with the deployment model and infrastructure you choose.

User authentication

SSO integration lets you centrally manage authentication and access management using your identity provider.

FAQ

Where is case data stored?

Only on your machines. Herfindahl is designed for local-first execution and controlled deployment models that keep case data fully under your organization’s control.

Can we restrict where inference happens?

Yes. Herfindahl supports fully customer-controlled routing for LLM inference, including restricting inference to the EU.

Is data from different cases handled separately?

Yes. Herfindahl is designed from the ground up to keep data from different cases completely separate, preventing cross-case leakage.

Do you train on customer data?

No. Herfindahl does not use your files, prompts, or outputs to train models. Under our customer-controlled deployment models, data remains entirely within your controlled environments and is never processed by us.

Serious about security?

Book a demo to see Herfindahl in action.