Privacy and training
The two questions people ask about an assistant reading live project data are: who else sees what I typed, and is it being used to train something. This is the answer to both, in plain words. The binding version is the AI Product Terms.
The short version
Your prompts and the answers are never used to train an AI model. Not by Teralo, and not by the AI provider behind it. That is a contractual restriction rather than a preference, and it applies to everything you type and everything the assistant returns.
You keep everything. You retain the rights and the intellectual property in what you put in and in what comes out. Teralo's licence to your input covers running the service, and explicitly does not extend to training models on it.
Using AI is optional. It is a permission on a project, and a project without it has no AI features at all.
Who can read your conversations
A conversation belongs to the person who started it. No screen in Teralo shows one person another person's chats: not a colleague, not the host organisation, not an organisation administrator. There is no shared inbox of conversations and no permission that grants access to somebody else's.
Deleting a conversation deletes its messages and any files you attached to it. Deleting the project takes its conversations with it.
The same privacy is what makes a conversation a poor record. If an answer matters to the job, put it somewhere the job can see it: export the conversation, or raise the observation, or send the mail. See AI Chat.
What leaves Teralo, and when
Nothing leaves until you ask a question. There is no background indexing of your project by an outside service.
When you do ask, what is sent to the model is your question, the conversation so far, and the results of the lookups the assistant chose to make: the rows it read, formatted as text. It is not given a copy of your project, and it is not given anything your own permissions do not already reach. See What the AI can read.
The model is Google Gemini
Every AI feature in Teralo runs on Google Gemini. There is no second provider and no per-feature choice of model.
Processing and data retention
Under Teralo's agreements with its AI provider, input and output are processed in-region and deleted on a defined retention period, typically 30 days or less:
- Australia, processed and retained only in Australia
- United States and North America, processed and retained only in the US
- Rest of world, delivered under equivalent privacy principles, not retained indefinitely
Retention there is the AI provider's, and it is separate from Teralo's own storage of the conversation, which lasts until you delete it.
What not to put in
The terms are specific about this, and it is worth reading before somebody pastes something they should not.
Do not submit sensitive personal information: health data, racial or ethnic origin, political opinions, religious beliefs, or sexual orientation. That includes pasting it in from a document.
More generally, you are responsible for what you put in, including whether you had the right to put it in. A drawing you are not licensed to share does not become shareable by going through an assistant.
What not to do with the answer
Verify it before you rely on it. The assistant can misread, can miss, and can be confidently wrong, which is why answers come with the records they were built from. Open one.
It is not professional advice. Nothing it produces is legal, financial, medical or engineering advice, whatever it sounds like.
Do not use it for automated decisions about people. Decisions with a legal or significant effect on somebody, employment among them, need a human making them.
Say so when you publish it. If you share AI-generated output beyond your own team, disclose that it was AI-generated, attribute it to you or your company, and review it first.
Feedback
Feedback you send about the AI features can be used to improve them, and is not treated as confidential. That is feedback specifically, and is separate from your prompts, which are not used for anything beyond answering you.
Where the full terms are
Everything here is a summary written to be readable. The full and binding version, including the disclaimers, the restrictions on use and the data-handling commitments, is the AI Product Terms. The Privacy Policy covers Teralo more broadly, and the Security page covers how the platform itself is protected.