1. Parties and scope
This agreement is between the organization whose owner turns on an AI feature (the “Controller”), acting through its owner, and Daniel Hrynusiw OSVČ, Bořetická 4133/6, 628 00 Brno-Židenice, Czech Republic, IČO 23347406 (the “Processor”, “we”). It adds to the Data Processing Agreement (“DPA”) for the two AI features of the Service: the Assistant and AI connections.
Everything the DPA says about organization data applies to the data these features handle, including breach notices, audits and help with requests. Where this agreement and the DPA differ on that data, this agreement prevails.
2. What the AI features are for
- The Assistant is an AI chat inside the Service, on the Enterprise plan. Members ask it about their time off, attendance and colleagues, and ask it to make changes. It reads only what the member asking can already see, and every change waits until that member confirms it.
- AI connections let a member connect an AI app they already use, such as Claude or ChatGPT, to their own account. The app works with that member’s own permissions in the organization.
Both are off until the owner turns them on, and the owner can turn either off at any time.
3. The data involved
- Chats: what a member writes to the Assistant, its answers, and the Service data it reads to answer, such as requests, balances, attendance and directory entries the member can see.
- The display record: the copy of each chat we keep so the member can read it back, with its tool activity lines, confirmation cards and their status, and messages the Assistant refused.
- Usage records: for each step of a chat, who, when, which model, the number of tokens and the cost. They hold no content.
- Connection records: for each AI connection, the app’s name and address, the member, the organization, the access granted and when it was last used.
The Assistant also receives the name and role of the member who is asking. Chats are organization data: we process them for the Controller under the DPA, on its instructions, and not for any purpose of our own.
4. The Controller’s instructions
The owner’s settings in the Service are the Controller’s documented instructions under the DPA: turning the Assistant on or off, and allowing AI connections as read only, read and write, or not at all. A change takes effect as an instruction when it is saved.
The Controller makes sure it has a legal basis for using the AI features with its people’s data, and tells its people about them as Article 13 GDPR requires. The “Tell members” option sends everyone a notice in the app and on their phone; it does not replace the Controller’s own information to its people.
For its own data protection impact assessment, the Controller finds the data flow, sub-processors, retention and safeguards on the DPIA information page.
5. Anthropic as a sub-processor
The Assistant runs on Anthropic, PBC, a company in the United States, which the Controller authorizes as a sub-processor by accepting this agreement. Anthropic receives chat data only from organizations that turn the Assistant on.
- Anthropic stores chats in the United States, and may run inference in any country where Anthropic runs its models.
- The transfer rests on the EU Standard Contractual Clauses, Module 3 (processor to processor). Anthropic is not certified under the EU-U.S. Data Privacy Framework.
- Anthropic never trains models on the Controller’s data, and neither do we.
- Anthropic keeps a message its safety systems flag for up to 2 years, to enforce its usage policy.
- Anthropic tells us about changes to its own sub-processors 15 days ahead. We pass each change on to the Controller as soon as we receive it, and the Controller may object as the DPA describes, or turn the Assistant off.
6. Retention and deletion
- We keep a chat’s display record for 180 days after its last message, and a member can delete their own chats sooner.
- Anthropic deletes a chat 30 days after its last message, or at once when the chat is deleted in the Service.
- When a member leaves the organization, is removed from it, or deletes their account, their chats in that organization are deleted at once, here and at Anthropic, and any change still waiting for their confirmation is cancelled.
- When the organization is deleted, every chat it holds is deleted, here and at Anthropic.
- Usage records are kept for 90 days. Each week then rolls into a total per person, kept for 180 days, and after that into a total for the organization that names nobody.
- Connection records are kept until the member disconnects the app.
- Turning the Assistant off stops new messages at once. Chats stay readable until they are deleted as above.
7. What the owner and org admins see
The owner and org admins see how much each person uses the Assistant and how many of their messages it refused. They never see what anyone wrote.
8. AI connections
Allowing AI connections is the Controller’s documented instruction to hand a member’s app what that member can see in the organization, and, with read and write, to let it make the changes the member could make. The company behind the app is the member’s choice, under the member’s own agreement with it. It is not our sub-processor, and we have no agreement with it.
Our responsibility ends when the data reaches the app. Personal accounts of AI apps may keep what they read or train on it, so the Controller’s own policy decides which apps its people may use.
9. Versions
We record who accepted this agreement, when, and which version. When we publish a new version, the owner is asked to accept it on their next visit, and the Assistant and AI connections keep working meanwhile. Until the owner accepts it, the version they last accepted applies.
10. Contact
Daniel Hrynusiw OSVČ
Bořetická 4133/6, 628 00 Brno-Židenice, Czech Republic
daniel@hrynusiw.cz