Legal Prompting - Professional secrecy and AI infrastructure

Welcome back to Legal Prompting, I'm Nicola Fabiano and this is episode 8.

In the previous episode, we saw how to integrate legal prompting into corporate compliance workflows

with structured processes, governance and traceability.

Today we tackle a topic that underpins everything,

professional secrecy and the choice of AI infrastructure.

The principle is simple but often underestimated.

When a lawyer, a legal consultant or a DPO enters information covered by professional secrecy into an AI system,

they are making a deontological choice, not just a technical one.

The infrastructure on which the model runs, the place where the data is processed,

the parties who can access it, the contractual guarantees in place, the applicable jurisdiction,

all of this is part of respecting that secrecy.

No well-written prompt can compensate for inadequate infrastructure.

Let's look at three concrete applications.

First application, analyzing a confidential file.

If I need to ask a model to summarize the documents of a dispute or to compare clauses of a contract covered by an NDA,

the first step is not writing the prompt.

It is verifying where that data will go, whether it will be used for training,

who will have access to the logs, how long it will be retained

and whether the provider offers guarantees compatible with the duty of confidentiality.

For highly sensitive data, a model run locally or on European infrastructure with a solid DPA

and explicit no training clauses is often the only coherent option.

Second application, privacy or health-related advice.

Special categories of data require reinforced caution.

Even when the use case appears generic,

a single identifying detail can turn the prompt into a processing of sensitive personal data.

The operational rule is preventative pseudonymization or, where possible, abstraction of the case.

If the provider does not offer adequate guarantees on extra EU transfers, sensitive data must not leave my perimeter.

Third application, drafting opinions on extraordinary transactions or criminal proceedings.

Here technical confidentiality is not enough.

One must consider the provider's jurisdiction, access requests from foreign authorities,

exposure to regulations such as the Cloud Act and the implications of Italian Law 132-2025

and the AI Act regarding infrastructure.

Three cross-cutting operational rules.

First rule, classify before prompting.

Every piece of information entering a model must be classified by confidentiality level.

Without classification there can be no informed choice of infrastructure

and every subsequent decision is blind.

Second rule, document the choice.

The client file must be able to show which tool was used, with what contractual guarantees,

on what legal basis and why that choice was proportional to the case.

Documentation protects the client and protects the professional.

Third rule, favor abstraction.

Where possible, replace identifying data with placeholders.

Work through categories and schemes.

Reconstruct the context only in your own mind.

The model helps you reason.

It does not need to know the identity of the persons involved.

Professional secrecy is not a constraint that limits the use of AI.

It is the framework that makes it legitimate.

Without this framework, every efficiency gain turns into a deontological and disciplinary risk

and every operational advantage becomes a latent liability.

In the next episode we will enter the heart of the AI Act,

which specific obligations fall on the legal professional,

how human oversight is articulated

and what transparency in the use of AI systems concretely means.

To explore these topics further

and receive weekly reflections on the relationship between law, privacy and technology,

I invite you to subscribe to the newsletter at nickfab.eu.

Thank you for listening!