Security and data

An autonomous system with credentials to your analytics and CMS needs clear boundaries. Agents limits what each connection can do and where it can be used.

​

Every integration is read-only except the CMS

Agents can’t change data through its read-only connections. The CMS connection is the only integration with write access. It writes completed work to your CMS as a draft. Nothing publishes automatically, so a person reviews each draft before it goes live.

​

You grant and revoke access through your own accounts

Google integrations connect over OAuth with limited scopes. You grant access from your own Google account and can revoke it there. API keys for third-party tools remain yours and can be rotated.

​

Writesonic is SOC 2 Type 2 and HIPAA compliant

Writesonic is SOC 2 Type 2 and HIPAA compliant.

​

Customer data stays out of model training

We don’t use customer data to train models.

​

Work runs across Anthropic and OpenAI

Agents routes model work across Anthropic and OpenAI. OpenAI models run on Azure. If one provider degrades, cross-provider fallback keeps work running.

​

Connections are scoped to specific projects

Connections belong to a workspace, then each connection is assigned to specific projects. A connection isn’t available to every project by default. Agents can use it only where it has been assigned.

​

What’s next