> ## Documentation Index > Fetch the complete documentation index at: https://cogno.studio/docs/llms.txt > Use this file to discover all available pages before exploring further. # Security overview > How Cogno stores, protects, processes and deletes your data — a reference for IT and security reviews. This page is for the people who approve Cogno for their company: IT, information security and procurement. It explains where your data lives, what runs on your own computers, how access is controlled, how AI processing works, and how data is deleted. For step-by-step setup of two-step verification and the audit log, see [Secure your workspace](https://cogno.studio/docs/security.md). In this page, **your data** means what your team enters or uploads to Cogno and what Cogno fetches from services you connect. ## Shared responsibility | Cogno (Genaxis, Inc.) is responsible for | Your organization is responsible for | | - | - | | Protecting the Cogno service and the infrastructure it runs on | Who your members are and which role each one has | | Protecting your data and credentials that Cogno stores | Which external services you connect, and what you allow them to share | | Keeping each Workspace's data and permissions separate | What your team enters into or uploads to Cogno | | Backups, monitoring and responding to security incidents | The computers and servers where you run the Cogno Engine | | | Your own agreements and settings with the AI providers your agents use | ## Where your data is stored | Data | Service | Location | | - | - | - | | Tasks, comments, documents, uploaded files, meeting recordings, application data | Google Cloud (database, object storage, application runtime) | Tokyo, Japan (`asia-northeast1`) | | Accounts, sign-in sessions, OAuth credentials for connected services | Supabase | Tokyo, Japan (`ap-northeast-1`) | | AI processing for Cogno's own features | Amazon Bedrock, Japan cross-Region inference | Tokyo and Osaka, Japan | Some supporting services process limited data outside Japan: web hosting access logs, error monitoring, product analytics, email delivery, billing and mobile push notifications. The [subprocessor list](https://cogno.studio/subprocessors) names each one, the data it handles and where. ## What runs on your computers and what runs in Cogno Cogno has two parts: | Part | Where it runs | What it does | | - | - | - | | Cogno service | Cogno's cloud (Google Cloud, Tokyo) | Stores your Workspace, Tasks and Memory; connects to your integrations; runs Cogno's own AI features. | | Cogno Engine | Your computer (with Cogno Desktop) or your own server | Runs coding agents such as Claude Code and Codex on folders on that machine, and reports the run's conversation, activity and results back to the Task. | The Engine connects out to Cogno over HTTPS; see [Use Cogno behind a corporate proxy](https://cogno.studio/docs/network-and-proxies.md) for the hosts to allow. A [Remote Server](https://cogno.studio/docs/remote-servers.md) accepts no inbound network connections — Cogno Desktop reaches it only through your SSH tunnel. The Engine signs in with its own machine credential, separate from any member's sign-in. Cogno stores only a hash of that credential. Removing a machine in **Settings** revokes its credential immediately. You also choose how much an agent may do on each machine — see [Control access during agent runs](https://cogno.studio/docs/execution-access.md). A Workspace can also use [Cloud](https://cogno.studio/docs/cloud-environments.md), a managed environment where Cogno keeps working while your computers are off. Cloud is optional. ## Integrations and permissions * Cogno reads from and acts in an external service only within the access you approve when you connect it. * Each Project chooses which connected sources and repositories it uses, and Project tool permissions cannot exceed the Workspace's. * Disconnecting a service stops new reads and actions in it. Data fetched from it is deleted on the schedule described for that integration. * Cogno does not use data from connected services for any AI model training, with or without consent. Each integration page lists the permissions it requests and what happens on disconnect — for example [Slack](https://cogno.studio/docs/slack-integration.md). See all integrations in [Integrations](https://cogno.studio/docs/integrations.md). ## Sign-in and access control * **Sign-in**: email and password, **Sign in with Google**, or **Sign in with Apple**. Passwords are stored only as hashes by the authentication service, never in plain text. * **Two-step verification**: any member can protect their account with an authenticator app. * **Required MFA**: on the Business plan, admins and Owners can require two-step verification for every member of the Workspace. * **Roles**: every member has one role — **Viewer**, **Member**, **Admin** or **Owner** — and every request is checked against it. Someone who is not a member of a Workspace cannot reach its data. * **API keys**: shown once at creation, stored only as hashes, and revocable at any time. See [Manage API keys](https://cogno.studio/docs/api-keys.md). * **Cogno's own changes**: Cogno changes settings only with the requesting member's permissions, and admins choose per group whether it may do so freely, must ask, or never. See [Secure your workspace](https://cogno.studio/docs/security.md) to set these up. ## AI processing Cogno uses generative AI in two ways: | | Cogno's AI features | Your coding agents | | - | - | - | | Provider | Amazon Bedrock, managed by Cogno | The provider you sign in with, such as Claude Code or Codex | | What is sent | The content a feature needs, such as Task text, documents and meeting transcripts | Whatever the work you give the agent requires | | Governing terms | Cogno's agreement with AWS | Your organization's agreement and settings with that provider | Cogno does not use your data to train or improve AI models. Under the Amazon Bedrock service terms, Bedrock inputs and outputs are not shared with the model provider and are not used to improve base models. Audio transcription for [Capture on mobile](https://cogno.studio/docs/mobile-capture.md), an experimental feature, uses Google Gemini outside Japan. The [subprocessor list](https://cogno.studio/subprocessors) describes this route. ## Encryption and secrets * Connections from your browser, apps and Engine to Cogno use TLS 1.2 or higher. * The application reaches the database only over encrypted connections. * The database, files and backups are encrypted at rest. * Private files are delivered through short-lived signed URLs. * Service secrets are kept in Google Cloud Secret Manager. * API keys and machine credentials are stored only as hashes. ## Retention and deletion | What | How long | | - | - | | Workspace data | While you use Cogno. After termination you can export for 30 days; Cogno then deletes it from production and backups within a target of 90 days. | | Data from a connected service | While the connection exists; deleted within a reasonable time after you disconnect. | | Meeting recordings | Deleted from Recall.ai after 7 days. The copy Cogno keeps in Tokyo is deleted after 90 days. | | Access logs | Up to two years. | | Data under a legal retention duty | For the period the law requires. | An Owner can delete a Workspace in **Settings → General → Delete workspace**. If the Workspace has a Business subscription, Cogno cancels it first; if the cancellation fails, nothing is deleted. Once deletion completes, the Workspace's data is removed from the primary database. Backups are not edited record by record; deleted data leaves them as older backups expire. To take your tasks with you first, see [Export tasks](https://cogno.studio/docs/export.md). ## Backup and recovery * The primary database runs as a high-availability pair across separate zones in the Tokyo region. * Cogno keeps the 21 most recent daily backups, stored in the Tokyo region. * Point-in-time recovery covers the last 7 days. * The database has no public IP address; it is reachable only from Cogno's private network. ## Audit log Cogno records privileged actions — role changes, API key creation and revocation, execution environment changes and more — with the time, who acted, the target and the outcome. Admins and Owners can read the log and download it as CSV in **Settings → Security → Audit log**. See [Review the audit log](https://cogno.studio/docs/security.md#review-the-audit-log). ## Incident response When Cogno detects a security incident, it confirms the scope, contains it, investigates the cause, recovers and prevents recurrence. If your data is affected, Cogno notifies you with the scope, cause and remedial measures, with a target of within 72 hours of discovery, and follows any legal reporting deadlines that apply. ## Report a security issue Email **[privacy@cogno.studio](mailto:privacy@cogno.studio)** with a description of the issue and the steps to reproduce it. The same address handles security questionnaires and requests for a data processing agreement. ## Subprocessors Cogno uses Google Cloud, Supabase and Amazon Web Services to run the service, plus providers for web hosting, monitoring, email, meeting recording and payments. Each receives only what its function needs. The full list, with the data each handles and where, is at [cogno.studio/subprocessors](https://cogno.studio/subprocessors). Certifications held by these providers belong to them; they are not certifications of Cogno. ## Legal documents * [Privacy Policy](https://cogno.studio/privacy) * [Terms of Service](https://cogno.studio/terms) * [Data Processing Agreement](https://cogno.studio/dpa) * [Subprocessor list](https://cogno.studio/subprocessors) ## Related docs * [Secure your workspace](https://cogno.studio/docs/security.md) — two-step verification, required MFA and the audit log. * [Control access during agent runs](https://cogno.studio/docs/execution-access.md) — what agents may do on each machine. * [Use Cogno behind a corporate proxy](https://cogno.studio/docs/network-and-proxies.md) — hosts to allow on a managed network. * [Connect to a remote server](https://cogno.studio/docs/remote-servers.md) — run work on your own server. * [Manage API keys](https://cogno.studio/docs/api-keys.md) — create and revoke keys. This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.