Answers to common questions about Canopy Coworker: how it processes your data, who at your firm can use it, and where to find Canopy’s security documentation.
Plans and Permissions: Firm admins have access by default, and other staff need a role that includes Coworker.
👉 Good to know: Coworker is in beta. Your firm’s use of it is covered by the Beta Tester Agreement your firm accepted in Canopy.
Why It Matters: Your firm holds sensitive client data — before letting an AI assistant work with that data, your partners and IT reviewers need to know where it’s processed, who can see it, and what your firm agreed to.
For Canopy’s SOC 2 report, pen test summary, subprocessor list and incident response documentation, start with the Canopy Trust Center.
Table of Contents
Data Processing and AI Models
Where does Coworker process our data? Is it sent to an outside AI company?
Coworker currently runs on AI models hosted inside Canopy's own AWS environment through Amazon Bedrock, so client data is not shared with the underlying model providers. AWS is the same cloud provider that hosts the rest of Canopy. These models currently operate on a zero data retention basis.
Which AI model does Coworker use?
Canopy chooses and updates the models as AI improves, and your data is currently processed in Canopy's environment whichever model is in use.
Does any data leave the U.S. when Coworker processes it?
No. Canopy does not currently transfer Coworker data outside the U.S. For details on inference and log regions, contact Canopy Support for more information.
Will our client data be used to train AI models?
Because Coworker runs on models hosted in Canopy's AWS environment, the model vendors don't store or train on your data. Your client data is not used to train foundation models or change model weights.
Canopy does log Coworker inputs, outputs and automated task activity, and reviews it for accuracy and to improve Coworker's performance.
Can we opt out of Canopy using our inputs and outputs to improve Coworker?
Canopy plans to offer a firm-level opt-out, and the process is still being finalized.
Who are Canopy's subprocessors?
The Canopy Trust Center lists them. For a list specific to the Coworker beta, contact Canopy Support.
Access, Permissions and Controls
Does Coworker respect each user's Canopy permissions?
Yes. Coworker works with the permissions of the person using it, including when it runs automated jobs or recommendations. It can only see and do what that user can already see and do in Canopy.
Can we control which staff have access to Coworker?
Yes. Admins grant Coworker through role settings, and firm admins have access by default. Canopy's built-in default roles don't include Coworker, so you must give staff access through a custom role.
Approving automated jobs is a separate permission. Users with basic Coworker access can create automated jobs, but those jobs wait for someone with approval permission.
Can we limit what Coworker is allowed to do?
Yes. Trust settings let your firm decide which actions run automatically and which need a person to review them first. Coworker also asks for explicit approval before it makes changes in a session that involves sensitive data.
Is there an audit log of what Coworker did?
An audit log and session history are available in a Coworker session. You can also view Coworker actions in the general Activity Log.
Is Coworker available through Canopy's public API?
No. Coworker isn't available through Canopy's public API.
How do we turn Coworker off?
Your firm can turn off Coworker at any time. For what happens to your data when you leave the beta, see the Beta Tester Agreement.
Compliance and Documentation
Can we get Canopy's SOC 2 report, pen test summary or incident response terms?
Yes. The Canopy Trust Center is where Canopy shares its security documentation, including breach notification terms.
Is Coworker in scope for Canopy's SOC 2 Type II?
Yes.
Does Canopy have cyber insurance?
Yes. Canopy maintains cyber insurance.
How long are Coworker logs kept, and who can access them?
Coworker logs stay inside Canopy's environment, and only authorized Canopy team members can access them. For retention terms, see the Beta Tester Agreement.
Beta Agreement and Terms
Do we have to notify our clients or get their consent to use Coworker?
Canopy can't provide legal advice. Consult your own legal and professional advisors about your notification and consent obligations. Your firm is responsible for understanding and meeting its legal and professional obligations when it uses Canopy.
If our data is secure, why does Coworker warn us not to enter SSNs or financial information?
The warning is about typing sensitive data directly into the Coworker task box. When Coworker reads data already stored in Canopy, it goes through Canopy's existing permissions and controls.
Who at our firm can accept the Coworker terms?
Someone with authority to bind your firm. The terms are accepted in Canopy, so no DocuSign is needed.
Related Articles:
Summary
Coworker runs on AI models inside Canopy's AWS environment, keeps your data in the U.S., doesn't train models on your client data, and can only see and do what the person using it can. For security documentation, start with the Canopy Trust Center, and consult your own advisors about client notice and consent.
Need Help?
Need help? Contact Support or ask Penny, our AI Support Bot, for assistance.