Next liveBuild Your Own Calendly · zoomwithchipp.comSave your seat →

Security at Chipp

How Chipp protects your data

Chipp runs the AI agents and software your team depends on. This page explains, in plain language, what we protect for every customer, what depends on your plan or settings, and what stays your call. Each statement here matches how the platform works today.

Trust and compliance

Audited, and open about the details

Our Trust Center is the canonical place for our security program. Confidential documents are shared only through its request-access process.

  • SOC 2 Type II

    An independent auditor examines our controls. Request the current SOC 2 Type II report through the Trust Center.

  • HIPAA support

    HIPAA mode on eligible plans, plus a Business Associate Agreement (BAA) that you sign with us separately. It is not on by default for every account.

  • Independent testing

    Third-party penetration testing, with the status of open findings published on this page.

  • Confidential documents

    The SOC 2 report, the Data Processing Agreement (DPA) and the BAA are available on request through the Trust Center.

Data protection

Encrypted on the wire and on disk

Your conversations, files and settings are encrypted in transit and at rest, and the credentials you connect get an extra layer from our application.

Every customer

Built in for everyone

  • TLS in transit

    Connections to Chipp, and Chipp's built-in connections to AI model providers, use TLS. Our API also tells browsers to use HTTPS only.

  • Encryption at rest

    Databases on Google Cloud SQL, files on Google Cloud Storage, and files on Cloudflare R2 are encrypted at rest by those providers.

  • Connected credentials encrypted by Chipp

    OAuth tokens and keys for the integrations you connect are encrypted by our application before they are stored.

  • Browser security headers

    The Chipp dashboard and API send strict security headers, including a content security policy, to limit what a malicious script or frame can do.

Infrastructure and platform protection

Proven cloud, watched all the time

Chipp runs on Google Cloud with Cloudflare at the edge. Automated checks and monitoring guard the platform before and after every change.

Every customer

Built in for everyone

  • Google Cloud in the United States

    The core platform runs on Google Kubernetes Engine and Cloud SQL in Google Cloud's us-central1 region in Iowa.

  • Cloudflare at the edge

    Cloudflare provides DNS, TLS and the edge proxy in front of Chipp's web traffic.

  • Tenant isolation

    Access checks run on our servers, scoped to your organization and workspace. Apps the software factory hosts run in separate containers as a non-root user, on a read-only file system, behind default-deny network rules.

  • Secrets management

    Platform secrets are kept in a managed secrets store and delivered to production at deploy time, not kept in our source code. Code changes are scanned for leaked secrets.

  • Monitoring and incident response

    Centralized logs, alert rules and uptime checks watch production, and alerts go to our engineering team. We write a postmortem after significant incidents.

  • Audit trails

    Sensitive account actions, such as membership and role changes, access grants, publishing, and sign-in policy changes, are written to an audit log.

  • Dependency security updates

    Dependabot and a daily dependency audit flag vulnerable packages. Patch and minor updates merge only after CI passes, and major upgrades are held for review.

  • Security scan on software factory changes

    Changes the software factory makes go through an automated scan for risky code patterns and leaked secrets before they merge, and new high-severity findings block the merge.

Authentication and access

The right people, and only them

Sign-in is passwordless for builders, sessions are protected, and every protected request is checked on our servers.

Every customer

Built in for everyone

  • Passwordless builder sign-in

    Builders sign in with a one-time email code, or with Google or Microsoft. There is no builder password to steal or reuse.

  • Secure sessions

    Sessions live in HttpOnly cookies that page scripts cannot read. You can sign out of one device or every device, and a deactivated account loses access on its next request.

  • Rate limits and progressive lockout

    Sign-in attempts are rate limited per account and per network, and repeated failures trigger a lockout that grows longer as failures continue.

  • Roles and scoped permissions

    Organizations have owner, admin and member roles. Workspaces have owner, editor and viewer roles. Protected actions are checked against server-side records, not against what the browser claims.

  • Scoped API keys

    API keys carry read or write scopes per resource and an optional expiry. We store only a hash of each key and show the key once.

Plan or setting

Depends on your plan or settings

Available when the condition on each item is met.

  • Your setting

    Two-factor authentication

    Add a passkey or an authenticator app, with recovery codes. Organization owners and admins can require it for every member at sign-in.

  • Admin configuration

    Single sign-on

    Owners and admins can connect your identity provider over OpenID Connect and require members to sign in through it.

  • Studio Pro and higher

    Member audit log

    Organization owners and admins can review a log of membership and permission changes.

  • Your app settings

    End-user sign-in for your agents

    You choose whether people must sign in to use an agent, and how: email code, password, Google, Microsoft and other supported providers.

AI and your data

Where your data goes when an agent answers

To answer a message, an agent has to send context to the AI model you picked. Here is what goes, and what decides where it goes.

  1. You pick the model

    Each agent runs on the model you select. That choice, together with your settings, decides which provider processes its messages.

  2. Chipp builds the context

    Each turn can include the prompt, recent conversation history, and relevant excerpts retrieved from your knowledge sources. Images and files a user shares in chat can also be included so the model can read them.

  3. The provider answers

    The context goes to that model's provider for inference. Depending on the model, the request goes to the provider directly or through a model routing service that also processes it.

  • Excerpts from knowledge sources

    For your knowledge sources, the model gets the matched excerpts it needs for a turn, not every document you uploaded.

  • Terms vary by provider and plan

    Retention and training terms depend on the model, the provider, your plan and your settings. We do not make one promise that covers every provider.

  • Embeddings for search

    To search your knowledge, the text of document chunks and search queries is sent to an embedding provider and turned into vectors.

  • Ask which providers apply

    Which providers see your data depends on the models and settings you choose. Our team can confirm the providers for your setup.

See our subprocessor list in the Trust Center (opens in a new tab). For the model providers behind your agents, ask our team.

Data lifecycle and privacy

Your data, from upload to deletion

How long data is kept depends on what it is, your plan and your settings. We do not apply one retention window to every kind of data.

  • Stored so your agents keep working

    Agents, knowledge sources and conversations are stored in your account so your agents can use them. You and your users can delete them. Zero data retention, where you turn it on, keeps chat content only for the active session.

  • Delete in the product

    You can delete conversations and agents in the product, and an organization owner can delete an organization once no other members remain.

  • Account deletion requests

    Ask for deletion of your whole Chipp account from the Chipp iPhone app or by contacting us. Our team processes each request, as described in our Privacy Policy.

  • Copies that expire on their own

    Some copies, such as logs and backups, are not removed at the moment of deletion. They expire on their own schedules.

Read our Privacy Policy and Cookie Policy, or contact us about a deletion request.

Plan and setting controls

Controls that depend on your plan or settings

These controls are not on for every customer. Each one names the plan, setting or agreement it depends on.

  • Studio Pro and higher, turned on by an admin

    Zero data retention

    When an organization turns it on, chat content is held in short-lived storage for the active session instead of being saved long term, and only models from supported providers run.

  • Studio Pro and higher, HIPAA mode on, signed BAA

    HIPAA mode and a BAA

    HIPAA mode adds technical guardrails, and a BAA covers the legal side. You need both, and neither makes a workflow compliant on its own. Details below.

  • Enterprise agreement

    Dedicated infrastructure and data residency

    Dedicated infrastructure and data residency needs are discussed as part of an enterprise agreement. They are not part of self-serve plans.

Plan and setting controls

HIPAA support in detail

Using Chipp does not by itself make your app or workflow HIPAA compliant. Compliance depends on your plan, your configuration, a signed BAA, and the safeguards you run on your side.

What has to be true

  1. An eligible plan

    HIPAA mode for Chipp agents is available on Studio Pro and higher plans.

  2. HIPAA mode turned on

    You turn it on for each agent, or for a whole workspace through its app defaults. Software factory projects have their own HIPAA switch, which the project owner turns on.

  3. A signed BAA

    A BAA with Chipp is a separate agreement. Request one through the Trust Center before any protected health information (PHI) enters Chipp.

  4. Your safeguards

    You decide what data goes in, who can reach it, and how your own people, systems and connected tools handle it.

What HIPAA mode enforces

Each control names where it applies: Chipp agents, software factory projects, or both.

  • Chipp agents

    Sign-in is required

    Turning on HIPAA mode makes every user of the agent sign in, and sign-in cannot be loosened while HIPAA mode stays on.

  • Chipp agents

    Shorter sessions with a warning

    End-user sessions on HIPAA agents are shorter and do not renew in the background. People see a warning before automatic sign-out.

  • Software factory projects

    Covered AI models only

    In a HIPAA software factory project, AI calls go only to covered models, and agents linked to it run only Anthropic (Claude) and Google (Gemini) models. Other models are refused.

  • Software factory projects

    Uncovered services blocked

    In a HIPAA software factory project, platform AI services without a BAA, such as embeddings, text to speech and web search, are blocked.

  • Software factory projects

    Analytics without personal data

    For HIPAA software factory apps, Chipp Insights drops emails, custom properties and city, and strips query strings from page paths.

  • Agents and projects

    Stays on once set

    HIPAA mode on a software factory project cannot be turned off. On agents, a workspace policy keeps it on for every agent it covers.

What stays with you

HIPAA is a shared responsibility. These parts are yours:

  • Sign a BAA with Chipp before PHI enters the platform.
  • Turn on HIPAA mode for every agent and project that handles PHI, and keep sign-in required.
  • Connect only tools and integrations that your own agreements cover for PHI, and keep patient data out of knowledge files.
  • Manage who has access: roles, two-factor sign-in, and removing people who leave.
  • Run your own HIPAA program: risk analysis, policies, workforce training and breach notification.

Penetration testing and open findings

Independent testing, published

We commission independent penetration tests of the Chipp platform. In the spirit of transparency, we publish the status of open findings here while we fix them. Detailed reports stay behind the Trust Center's request-access process.

Latest assessment: June 24-30, 2026.

Need the full reports? Ask through the Trust Center request-access process (opens in a new tab). Found an issue? Email security@chipp.ai.

Doing a security review?

Our Trust Center has our subprocessor list and takes requests for the SOC 2 Type II report, the DPA and the BAA. Our team can answer the rest.