·6 min read
Where does my company's data go when I use AI?
The question every manager asks before adopting AI. See the difference between pasting information into an AI chat and having a brain on a server dedicated to your company.
Before talking about productivity, almost every manager asks the same question: "what about my data, where does it go?". It's the right question. Contracts, figures, client conversations and internal decisions are the most sensitive things a company has.
In this article we explain clearly what happens to that information in two scenarios: when the team pastes data into a general-purpose AI chat such as ChatGPT, and when the company uses a brain like Kortia.
1. AI is already in your company, even if nobody decided it
In many companies, AI arrived without any project: someone pastes a contract to get a summary, someone else pastes a spreadsheet to build a report, a salesperson pastes a client's history to write an email.
Each of these actions looks small. Added up, they mean company information is leaving for tools the company didn't choose and doesn't control.
2. What happens when someone pastes data into an AI chat
Tools like ChatGPT are very useful, and there are business plans with stricter rules. The risk lies in loose use, on personal accounts, without the company knowing. In that scenario:
- The information leaves the company and is stored in the account of whoever pasted it, not somewhere the company controls.
- Nobody knows what was sent, by whom or when.
- Depending on the plan and settings, conversations may be used to improve the models. On ChatGPT's individual plans this happens by default unless the user turns it off.
- When that person leaves the company, the history leaves with their account.
3. At Kortia, every company has its own server
Kortia works differently. Each client has its own instance on a dedicated server. The company's memory (notes, documents, meeting records and the history of conversations with the brain) is stored there, isolated.
One company's data never mixes with another's, and the knowledge lives somewhere that belongs to the company, not in each employee's personal account.
4. What happens when you ask the brain a question
Transparency also means saying what leaves the server. To answer, Kortia uses an established language model, Anthropic's Claude. The path is:
- The agent reads, on your company's server, only the parts of the memory needed for that question.
- The question and those excerpts are sent to the model through Anthropic's business API, with a key that belongs to your company alone.
- Under Anthropic's commercial terms, data sent through the API is not used to train its models.
- The answer comes back and is recorded on your company's server, not in a personal account.
5. Agents with clear limits
Kortia's agents work within fixed rules that don't depend on each user's settings:
- They read only your company's brain and sources.
- They don't browse the internet, run commands or access other files on the server.
- They only write to the memory once a person confirms.
6. Your knowledge in an open format
The brain's memory is stored as text notes, in an open, readable format. Your company's knowledge isn't locked into a closed format that only one tool can read.
7. Questions to ask any AI vendor
Whatever solution you evaluate, it's worth asking:
- Where is my company's data stored?
- Is it kept separate from other clients' data?
- What is sent outside, and to whom?
- Is my data used to train models?
- What can the agents read and do on their own?
- In what format is the knowledge stored?
Using AI safely isn't about avoiding the technology; it's about knowing exactly where the data goes. With a dedicated server, clear rules for agents and transparency about what's sent outside, your company gets the benefits of AI without losing control of its most valuable asset: its own knowledge.