Lini – Frequently Asked Questions about our AI Engine

As AI becomes a core part of the DiliTrust Suite, you may have questions about how it works, where your data goes, and what guarantees you have. This article answers the most common questions from our legal teams about Lini, the AI built into your DiliTrust platform.

⭐ What Lini is

Does DiliTrust build its own language model?

No. DiliTrust builds the legal intelligence layer around leading language models. We do not claim to build the underlying LLM (the large language model that generates the text).

DiliTrust develops the layer that makes AI useful for legal and governance work: legal data models, orchestration, preprocessing and post-processing, retrieval, playbooks, extraction logic, workflow integration and the controls around AI-assisted output. That layer prepares the right context, applies legal logic and playbooks, controls how the result is returned, and keeps the workflow inside the DiliTrust application.

What does DiliTrust mean by "proprietary"?

Our proprietary work is the legal intelligence around the model: the data structures, orchestration, context handling, playbooks, processing logic and product controls that help teams work with contracts, matters, entities and governance records. It is not a claim that DiliTrust trains or owns the underlying foundation model.

✅ Data & Privacy

Does DiliTrust use our data to train its AI?

No. No model training is performed by DiliTrust without the customer's consent. Our model partners do not use customer data to train their models. Customer data remains in the customer environment and is not shared across clients.

Customer information is used only to answer the requested AI task. It is not used to train a shared model.

Where are our documents stored?

In both configurations described below, customer documents remain stored on DiliTrust infrastructure.

• DiliTrust-managed frontier configuration: a relevant part of the AI processing is performed by the configured partner model provider, and the response is returned to DiliTrust. The provider does not use customer data to train its model and does not store the customer data under the documented setup.
• Sovereign configuration: processing is carried out within the DiliTrust perimeter under the applicable sovereign setup.

In CLM, does the AI learn from the labelled attributes I configure, including tenant-specific ones?

Customer-labelled attributes and asset descriptions can guide AI tasks, including extraction and other contextual actions in the relevant DiliTrust workflows. This is task guidance, not model training.

In practice, when you label attributes in your CLM tenant, those examples and their descriptions (hints) are used as in-context guidance to help the AI recognise similar notions in new contracts, within your own tenant only.

⚡ AI Technology & Architecture

Is DiliTrust's AI generative or extractive?

Today, DiliTrust's base AI is generative by nature. However, it is used for extraction purposes on certain features, so depending on the use case it can also be considered extractive.

Which language models does Lini use?

The underlying language model is selected from leading models available on the market. DiliTrust does not present these models as its own. There are two ways to run AI:

• DiliTrust-managed frontier configuration: a leading frontier model selected for quality and fit. Examples may include GPT, Claude or Gemini families, subject to the customer environment. Best fit for customers who want access to leading frontier models through the standard DiliTrust experience.
• Sovereign configuration: an open-weight model operated within the DiliTrust perimeter, for customers with specific sovereignty, residency or contractual requirements.

In both configurations, DiliTrust's legal intelligence layer, orchestration and application controls remain in place. DiliTrust remains responsible for the legal intelligence layer, application integration, access control and audit trail.

The right configuration depends on your performance, security, data residency and contractual requirements. Your DiliTrust account manager can confirm the configuration, provider, version, region and data path for your environment.

Can we connect our own language model?

If your organisation already uses a language model approved by its own teams, it can be connected through Bring Your Own LLM (BYOLLM). See Suite – Bring Your own LLM to DiliTrust for the connection requirements. Please reach out to your DiliTrust account manager to review the configuration best suited to your requirements.

⚖️ Third-Party AI Providers & Responsibility

What happens if I choose to connect my own AI provider instead of the configuration managed by DiliTrust?

DiliTrust is committed to end-to-end security within its own infrastructure. However, if you choose to connect an external AI provider under your own account:

• DiliTrust cannot be held responsible for where your data may transit once it leaves our secure environment.
• This option may be useful to benefit from the performance of larger or more specialised models.
• Before proceeding, ensure your legal and compliance teams have reviewed the data processing terms of the chosen provider.

Always consult your DiliTrust account manager before configuring a third-party AI provider.

✋ Security, Oversight & Certifications

How is access to the AI controlled?

AI functionality is integrated into the DiliTrust application and follows its authentication and authorisation mechanisms. Administrative access is restricted, and actions are logged across the platform and hosting infrastructure.

How are model changes managed?

DiliTrust manages code and model changes through a change-management process. Unit, integration and non-regression testing are performed as the code or model evolves. Human supervision checks output quality and non-regression.

Do we still need to review what the AI produces?

Yes. AI output is a working result, not an automatic decision. You must review and validate AI-assisted documents, summaries, minutes and recommendations before relying on them or making a binding decision.

Which certifications cover the AI?

The DiliTrust solution as a whole is certified to ISO 27001, ISO 27701 and SOC 2. There is currently no certification specific to the AI solution. These certifications do not replace a customer-specific review of the provider, data path, region or contractual commitments. DiliTrust does not claim automatic compliance with the EU AI Act.

❓ Still have questions?

For security, privacy, procurement or compliance reviews, contact your DiliTrust account manager or Customer Care at support@dilitrust.com. DiliTrust will confirm the provider, version, region, data path, retention position and applicable contractual details for your environment.