AI That Cannot Leave The Room
Ansarada AI runs inside your data room and nothing it touches leaves. Customer document content is never used to train any model. Ansarada holds enterprise contracts with its two LLM providers containing provisions that no customer data is retained, logged or used for model training. Every API interaction is stateless. AiDA answers only from content the user is already permitted to see.
The question every dealmaker is now asking
A partner pastes a draft SPA into a public chatbot to summarise it. It takes four seconds. It is the single most common confidentiality failure in dealmaking today and almost nobody reports it, because the moment it happens there is nothing to report. The document is gone. There is no log. There is no way to prove where it went or who saw it.
Ansarada built AiDA so that the useful thing and the dangerous thing are no longer the same action. You get the summary. The document never leaves the room.
How AiDA works
AiDA is Ansarada's Intelligent Dynamic Assistant, built for dealmaking rather than adapted to it. Four properties define how it handles your data.
It runs inside your data room. Only the specific, permission filtered content needed to answer a question is used for inference. Nothing leaves the room the way it would if a document were uploaded to a public chatbot.
It draws only on room content. AiDA cannot pad an answer with information from the open internet. That containment is also what keeps hallucination risk low. An assistant that can only see your documents can only answer from your documents.
It respects permissions already set. AiDA answers within the boundaries of a user's room role, Q&A role and team permissions. It cannot surface a document to someone who is not entitled to open it.
It is read only. AiDA finds and explains. It does not act, edit, move, share or delete.
Data retention and training, stated plainly
Ansarada uses a multi model architecture drawing on two trusted enterprise grade providers, with models selected for their specific use case.
The commercial terms behind that architecture matter more than the architecture itself. Ansarada maintains enterprise contracts with these providers that include provisions ensuring no customer data is retained, logged, or used for model training. All API interactions are stateless and governed by zero data retention commitments at the provider level.
The model is never trained on your prompts, your uploads, or your outputs.
Where AI processed content is stored
AiDA respects regional data sovereignty in line with each customer's configured data room environment. Document data and associated AI processed content, including OCR outputs, translated versions and extracted entities, is stored in the customer's appointed region.
Ansarada offers 14 storage locations across 13 countries: Australia (Sydney), Canada (Montreal), Germany (Frankfurt), Indonesia (Jakarta), Ireland (Dublin), Norway (Oslo), Saudi Arabia (Jeddah), Singapore, South Africa (Cape Town and Johannesburg), Switzerland (Zurich), United Arab Emirates (Dubai), United Kingdom (London), United States (California).
The location is selected before the room is created and fixed at that point. It cannot be changed afterwards. For a regulated transaction that is the correct behaviour. A jurisdiction that can be moved mid process is not a jurisdiction, it is a preference.
Which features use AI models
| Feature | Uses AI models |
|---|---|
| AI-Redact | Yes |
| AI-Translate | Yes |
| Bidder Engagement Score | Yes |
| Ask AiDA | Yes |