> For the complete documentation index, see [llms.txt](https://docs.maiagent.ai/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.maiagent.ai/maiagent-user-guide/en/org/security-and-privacy.md).

# Security and Privacy

Security and privacy highlights of MaiAgent for IT and security contacts: ISO 27001/27701 certification, encryption in transit, encrypted credential storage, role permissions, SSO, the activity log, H

This page summarizes the information security and privacy mechanisms MaiAgent provides, so that IT and security contacts evaluating adoption can grasp the key points quickly. You can also forward it directly to your company's security team. Each section links to the detailed instruction page in this manual.

{% hint style="info" %}
This page only lists mechanisms the platform already provides. For security questionnaire responses, certification documents, security incident reporting procedures, or contract terms, please contact your MaiAgent representative.
{% endhint %}

## At a Glance <a href="#at-a-glance" id="at-a-glance"></a>

| Area                 | Mechanisms MaiAgent provides                                                        | Details                                                           |
| -------------------- | ----------------------------------------------------------------------------------- | ----------------------------------------------------------------- |
| Certification        | Dual ISO 27001 and ISO 27701 certification                                          | [Certifications](#certifications)                                 |
| Transmission         | Site-wide HTTPS, with HSTS requiring browsers to use encrypted connections only     | [Encryption](#encryption)                                         |
| Credential storage   | Passwords, keys, and tokens used to integrate external systems are stored encrypted | [Encryption](#credential-encryption)                              |
| Who can use what     | Role permissions, resource assignment, model access permissions                     | [Access Control](#access-control)                                 |
| Sign-in              | Enterprise single sign-on (SSO) and LDAP identity sync                              | [Sign-In and Account Management](#sign-in-and-accounts)           |
| Auditing             | The activity log is append-only and cannot be edited or deleted                     | [Activity Log](#activity-log)                                     |
| Sensitive content    | Hooks mask PII and block malicious or inappropriate content                         | [Hook](#hook)                                                     |
| Conversation content | Agent monitoring records every AI reply; role permissions decide who can see it     | [Differences Between the Three Logs](#monitoring-vs-activity-log) |

## Certifications <a href="#certifications" id="certifications"></a>

MaiAgent holds dual certification for the **ISO 27001 Information Security Management System** and the **ISO 27701 Privacy Information Management System**. To obtain certification documents, please contact your MaiAgent representative.

## Encryption <a href="#encryption" id="encryption"></a>

### Encryption in Transit <a href="#encryption-in-transit" id="encryption-in-transit"></a>

* All connections to the platform use HTTPS; if you connect over HTTP, the platform redirects you to HTTPS.
* The platform uses HSTS to require browsers to use encrypted connections from then on, preventing connections from being downgraded to unencrypted HTTP.

### Encryption at Rest <a href="#encryption-at-rest" id="encryption-at-rest"></a>

* For SaaS, the database and uploaded files are encrypted at rest. See the tech manual [AI Gateway Security: Who Sees Prompts](https://docs.maiagent.ai/tech/ai-gateway/ai-gateway-security#q3-who-sees-prompts).
* For private cloud and on-premises deployments, the database and files are stored on your company's servers or storage devices. Encryption at rest is handled by your company's disk or storage layer; the platform itself does not additionally encrypt files.

### Encrypted Storage of Credential Fields <a href="#credential-encryption" id="credential-encryption"></a>

When integrating external systems, the platform needs to store their passwords or keys. The following fields are encrypted with AES-256-GCM before being written to the database:

* The service account password for [Identity Sync (LDAP)](https://docs.maiagent.ai/maiagent-user-guide/en/org/identity-sync.md)
* Secrets and access tokens used to integrate messaging platforms, such as LINE, Messenger, Instagram, WhatsApp, Microsoft Teams, LINE WORKS, and the mailbox password for Email platforms
* OAuth authorization information for connectors and MCP tools (client secret, access token, and refresh token)
* Provider credentials for [AI Gateway](https://docs.maiagent.ai/maiagent-user-guide/en/build/ai-gateway.md) bring-your-own-key (BYOK)

## Access Control: Who Can See and Use What <a href="#access-control" id="access-control"></a>

* **Roles**: A member's final permissions are the union of the default role and all custom roles. Feature permissions determine which features appear in the left menu; resource assignment (Agents, knowledge bases, messaging platforms, skills, and so on) combined with <mark style="color:blue;">Read</mark> permission determines which resources a member can see. Organization owners automatically have all permissions and resources. See [Role Permission Management](https://docs.maiagent.ai/maiagent-user-guide/en/org/roles/role-permission.md) and [Role Resource Visibility Control](https://docs.maiagent.ai/maiagent-user-guide/en/org/roles/group-resource-visibility.md).
* **Organization isolation**: Data is isolated per organization, and members can only access data from the organization they belong to.
* **Model access permissions**: Organization owners can disable AI models that are not allowed; disabled models do not appear in any model menu. See [Model Access Permissions](https://docs.maiagent.ai/maiagent-user-guide/en/org/model-access.md).
* **API Key**: Each key is bound to one member and one organization, and calls carry the same permissions as that member. Keys can be disabled or revoked at any time. See [API Key Management](https://docs.maiagent.ai/maiagent-user-guide/en/org/developer/api-keys.md).

{% hint style="warning" %}
**The default role applies to all members.** We recommend giving the default role only minimal permissions and assigning other permissions through custom roles. If you have set permissions but a member cannot see something, or someone who should not see it can, troubleshoot with the [item-by-item checklist](https://docs.maiagent.ai/maiagent-user-guide/en/org/roles/role-permission.md#faq-member-cannot-see-agent).
{% endhint %}

## Sign-In and Account Management <a href="#sign-in-and-accounts" id="sign-in-and-accounts"></a>

* **Enterprise single sign-on (SSO)**: Supports SAML (including ADFS), EIP, Keycloak, and Cloudflare Access. Employees sign in with their company accounts and do not need to remember a separate MaiAgent password. See [Third-Party Login (SSO)](https://docs.maiagent.ai/maiagent-user-guide/en/org/sso.md).
* **Identity sync (LDAP)**: Automatically syncs accounts and groups from directories such as AD, JumpCloud, and OpenLDAP. Accounts that are disabled in the directory or fall outside the sync scope are disabled in MaiAgent at the next sync, and the member's API Keys are deleted as well. See [Identity Sync (LDAP)](https://docs.maiagent.ai/maiagent-user-guide/en/org/identity-sync.md#faq-account-keeps-disabled).

Both of the above are Enterprise features.

## Auditing: Activity Log <a href="#activity-log" id="activity-log"></a>

The [Activity Log](https://docs.maiagent.ai/maiagent-user-guide/en/org/activity-log.md) records every creation, update, and deletion of data in your organization, as well as member logins. Each record includes the operation time, the member who performed the operation, the source IP address, and the changes made. Records can be exported to Excel by period for retention.

The activity log is append-only: the back office has no function to delete or edit records, the database layer also blocks modification and deletion, and even organization owners cannot remove any record. See [Activity Log: Data Retention](https://docs.maiagent.ai/maiagent-user-guide/en/org/activity-log.md#data-retention).

## Sensitive Content: Hook Message Interception <a href="#hook" id="hook"></a>

[Hooks](https://docs.maiagent.ai/maiagent-user-guide/en/build/hook.md) automatically step in before content is handed to the AI and before the AI's reply is sent. They are enforced by the system rather than relying on the AI's self-discipline:

* **Mask PII**: Masks phone numbers, emails, national ID numbers, credit card numbers, as well as names, addresses, and other personal data before passing them to the model.
* **Block content**: Blocks prompt injection attacks, inappropriate content such as violence or hate, and messages containing banned words.
* **Covers everything the AI reads**: Besides what users type, attachments, knowledge base retrieval passages, and results returned by tools also pass through the same Hooks.

Every block, rewrite, and pass-through is kept in the [Hook execution logs](https://docs.maiagent.ai/maiagent-user-guide/en/build/hook/execution-logs.md). By default, the logs keep only the outcome and not the original content before masking; see [Controlling Whether Original Messages Are Logged](https://docs.maiagent.ai/maiagent-user-guide/en/build/hook/execution-logs.md#log-privacy-toggle). The Enterprise plan includes Hooks automatically; for other plans, MaiAgent can enable them for your organization individually.

## Differences Between Agent Monitoring, the Activity Log, and Hook Execution Logs <a href="#monitoring-vs-activity-log" id="monitoring-vs-activity-log"></a>

Each of the three logs serves a different purpose, and security audits often require viewing them together:

|                         | Activity Log                                                                                           | Agent Monitoring (conversation records)                                                                                                                                                                                                                                          | Hook Execution Logs                                                                                                                                                                                                                                                                                                                                |
| ----------------------- | ------------------------------------------------------------------------------------------------------ | -------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | -------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
| What it records         | Data creation, updates, deletion, and member logins: who, when, from which IP, and what changed        | Every AI reply: user message, AI reply, Agent, model, response time, token usage, quality score, and the member who initiated it                                                                                                                                                 | Every Hook block, rewrite, and pass-through: which message, which Hook, and what action was taken                                                                                                                                                                                                                                                  |
| What it does not record | Conversation content, and simply opening pages or viewing data                                         | Setting changes and logins                                                                                                                                                                                                                                                       | Conversations of Agents with no Hook attached                                                                                                                                                                                                                                                                                                      |
| Who can see it          | Roles with <mark style="color:blue;">Activity Log Permission</mark>                                    | Organization owners see everything; other members need <mark style="color:blue;">AI Assistant Monitoring Permission</mark> and can only see Agents assigned to their roles. MaiGPT is assigned to the default role by default, so members' MaiGPT conversations are also visible | Roles with <mark style="color:blue;">Hook Permission</mark>; organization owners see everything, while other members only see records for Agents assigned to their roles. Records of colleagues' MaiGPT conversations are visible only to organization owners and members with <mark style="color:blue;">AI Assistant Monitoring Permission</mark> |
| Location                | <mark style="color:blue;">Organization Settings</mark> → <mark style="color:blue;">Activity Log</mark> | <mark style="color:blue;">AgentOps</mark> → <mark style="color:blue;">AI Assistant Monitoring</mark>                                                                                                                                                                             | <mark style="color:blue;">AgentOps</mark> → <mark style="color:blue;">Hook Execution Logs</mark>                                                                                                                                                                                                                                                   |
| Details                 | [Activity Log](https://docs.maiagent.ai/maiagent-user-guide/en/org/activity-log.md)                    | [Agent Monitoring](https://docs.maiagent.ai/maiagent-user-guide/en/build/agent-ops/evaluations.md#ai-agent-monitoring), [Viewing MaiGPT Conversations](https://docs.maiagent.ai/maiagent-user-guide/en/build/agent-ops/evaluations.md#view-maigpt-conversations)                 | [Hook Execution Logs](https://docs.maiagent.ai/maiagent-user-guide/en/build/hook/execution-logs.md)                                                                                                                                                                                                                                                |

Requests that call MaiAgent through the API are recorded separately in [API Call Logs](https://docs.maiagent.ai/maiagent-user-guide/en/org/developer/api-logs.md).

{% hint style="info" %}
**Conversation content is visible in two places in the back office:** <mark style="color:blue;">AI Assistant Monitoring</mark> (requires <mark style="color:blue;">AI Assistant Monitoring Permission</mark>), and <mark style="color:blue;">Customer Service Conversations</mark> → <mark style="color:blue;">All Conversations</mark> (requires <mark style="color:blue;">All Conversations Permission</mark>; the scope is messaging platforms assigned to the member's role with <mark style="color:blue;">View All Conversations</mark> checked, plus conversations assigned to the member). Organization owners can see everything in both places. Before adoption, we recommend deciding which roles need these two permissions and informing your members. For the member-facing explanation, you can cite [MaiGPT Security and Privacy](https://docs.maiagent.ai/maiagent-user-guide/en/work/maigpt/faq/security.md#admin-access-to-chat) directly.
{% endhint %}

## Mechanisms Not Currently Provided <a href="#not-provided" id="not-provided"></a>

Items that security questionnaires often ask about but the platform does not currently offer are listed here to avoid back-and-forth during evaluation:

| Item                                                                                               | Current status                                                                                                                                                                                                             | Alternative                                                                                                                                                                                                                                                                                                                                              |
| -------------------------------------------------------------------------------------------------- | -------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | -------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
| Privileged access management (PAM), just-in-time (JIT) authorization, privileged session recording | Not provided, and no integration with PAM products                                                                                                                                                                         | Use [role permissions](https://docs.maiagent.ai/maiagent-user-guide/en/org/roles/role-permission.md) to control who can use what, and SSO/LDAP so accounts are provisioned and disabled along with your company directory; administrative operations are kept in the [activity log](https://docs.maiagent.ai/maiagent-user-guide/en/org/activity-log.md) |
| Integration with external secret vaults (such as HashiCorp Vault or cloud Secrets Managers)        | Not provided. Passwords and keys for integrating external systems are encrypted by the platform with AES-256-GCM and stored in the platform database; see [Encrypted Storage of Credential Fields](#credential-encryption) | —                                                                                                                                                                                                                                                                                                                                                        |
| Automatic API Key rotation on every request                                                        | Not provided. An API Key remains valid until revoked, and an expiration date can be set                                                                                                                                    | Periodically create a new key in [API Key Management](https://docs.maiagent.ai/maiagent-user-guide/en/org/developer/api-keys.md), update the caller, and then revoke the old key                                                                                                                                                                         |
| Real-time delivery of the activity log or conversation records to a SIEM                           | Push and syslog output are not provided                                                                                                                                                                                    | [Export the activity log](https://docs.maiagent.ai/maiagent-user-guide/en/org/activity-log.md) to Excel by period, then import it into your company's systems                                                                                                                                                                                            |
| Integration with third-party runtime prompt injection protection products (such as AI firewalls)   | Not integrated                                                                                                                                                                                                             | Use the prompt injection and inappropriate content detection of [Hooks](https://docs.maiagent.ai/maiagent-user-guide/en/build/hook.md), or place the protection product at the network layer between users and MaiAgent; see the tech manual [AI Guardrails](https://docs.maiagent.ai/tech/advanced-genai-tech/guardrails)                               |

## Data Use <a href="#data-use" id="data-use"></a>

The AI models used by MaiGPT do not use conversation data for training; see [MaiGPT Security and Privacy](https://docs.maiagent.ai/maiagent-user-guide/en/work/maigpt/faq/security.md#ai-training-on-chat). For questions about each model provider's data processing terms, data storage, and retention, please contact your MaiAgent representative.

If your organization requires that knowledge base index data stay in an environment you manage, you can connect the knowledge base to your organization's own vector database (an Enterprise feature); see [How to Create a Knowledge Base: Vector Database](https://docs.maiagent.ai/maiagent-user-guide/en/build/knowledge-base/km-basic-settings.md#vector-database).

## Recommended Setup Before Adoption <a href="#recommended-setup" id="recommended-setup"></a>

1. Enable enterprise SSO or LDAP identity sync so accounts are provisioned and disabled along with your company directory.
2. Reduce the default role's permissions to the minimum, and assign other permissions through custom roles.
3. Decide which roles get <mark style="color:blue;">AI Assistant Monitoring Permission</mark> and <mark style="color:blue;">Activity Log Permission</mark>, and review them regularly.
4. For Agents that serve external users or handle personal data (for example, [image vision applications](https://docs.maiagent.ai/maiagent-user-guide/en/work/image-vision.md) that recognize ID documents or invoices), attach Hooks for PII masking and prompt injection protection.
5. Export the activity log by period on a regular basis for audit retention.

## Related Pages <a href="#related" id="related"></a>

* [Roles and Permissions](https://docs.maiagent.ai/maiagent-user-guide/en/org/roles.md)
* [Third-Party Login (SSO)](https://docs.maiagent.ai/maiagent-user-guide/en/org/sso.md)
* [Activity Log](https://docs.maiagent.ai/maiagent-user-guide/en/org/activity-log.md)
* [Hook Message Interception](https://docs.maiagent.ai/maiagent-user-guide/en/build/hook.md)
* [Automated Evaluation and Agent Monitoring](https://docs.maiagent.ai/maiagent-user-guide/en/build/agent-ops/evaluations.md)
* [MaiGPT Security and Privacy](https://docs.maiagent.ai/maiagent-user-guide/en/work/maigpt/faq/security.md)


---

# Agent Instructions
This documentation is published with GitBook. GitBook is the documentation platform designed so that both humans and AI agents can read, navigate, and reason over technical content effectively. Learn more at gitbook.com.

## Querying This Documentation
If you need additional information that is not directly available in this page, you can query the documentation by asking a question.

Perform an HTTP GET request on the following URL with the `ask` and `goal` query parameters:

```
GET https://docs.maiagent.ai/maiagent-user-guide/en/org/security-and-privacy.md?ask=<question>&goal=<user_goal>
```

`ask` is the immediate question: it should be specific, self-contained, and written in natural language.
`goal` is what the user is ultimately trying to achieve, the reason they need the answer. Sharing it helps GitBook give you a better, more relevant answer. A goal is most helpful when it describes the outcome the user wants rather than restating the question. For example, with `ask=how do I create an API token`, a goal like `automate deployments from our CI pipeline` lets GitBook tailor the answer to that use case.

The response will contain a direct answer to the question and relevant excerpts and sources from the documentation.

Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.
