News in Practice
PROFESSIONAL ETHICS · CONFIDENTIALITY
Legal copilots and professional secrecy: what a firm commits to when it deploys
Your vendor contract contains a confidentiality clause, and it is well drafted. It does not, however, answer the question professional secrecy raises.
For firms deploying a legal copilot on matters covered by professional secrecy, and for legal departments bound by an equivalent duty of confidentiality. What a clause organises, what it cannot cover, and what is verified instead.
The contract you signed to deploy a legal copilot contains a confidentiality undertaking, a prohibition on using your data for other purposes, a liquidated damages clause and a liability regime. It was reviewed by someone who knows contract law, and it is sound. The question to ask is not whether it is well drafted, but whether it addresses what professional secrecy requires. It does not, and that distinction is not theoretical.
What a confidentiality clause organises
It must be said without condescension, because the reasoning that follows is sometimes presented as a disqualification of contracts, which it is not: a serious undertaking with a serious provider is real protection, and the absence of one would be a failing.
A confidentiality clause is a liability mechanism: it organises the remedy for a breach. That is its function, it performs it correctly, and it deals effectively with almost every contractual risk a firm encounters.
It produces three useful effects: it deters, by attaching a cost to breach; it compensates, if breach occurs; and it permits termination. All three presuppose that the breach has occurred and can be established.
What professional secrecy requires
This is not an obligation of result accompanied by compensation. A breach of secrecy is, in itself, the harm: it is not a financial loss that an indemnity could make good, it is the rupture of what underpins the relationship between a lawyer and a client.
The practical consequence is direct. A firm whose privileged information has circulated is not restored by a liquidated damages clause. The provider pays, the firm is compensated, and nothing that occurred is undone, neither for the client nor for the lawyer as regards their duty.
The duty falls, moreover, on the lawyer and not on the provider: a breach engages the firm before its bar and before its client, whatever the contractual liability of the third party involved. Recourse against the provider is a separate and subsequent question.
A clause organises what happens after a disclosure. Professional secrecy concerns what must not happen.
Why the two do not overlap
To the difference in nature is added an evidential difficulty that is rarely discussed. To invoke a clause, one must establish that a disclosure occurred, which presupposes discovering it.
Yet the circulation of information is almost never detectable from outside. A firm that learns of a leak learns of it through its consequences, often months later, and without being able to trace its source among the many hands through which the matter passed.
Protection whose activation depends on detecting the very event it sanctions therefore offers less than its drafting suggests. That is not a defect of the clause: it is the limit of the contractual approach applied to this kind of duty.
The question of downstream sub-processors
One particular difficulty escapes contractual reasoning entirely, and it is systematic: the provider you contract with almost never performs alone.
It relies on a hosting provider, often on a model provider, sometimes on infrastructure or monitoring vendors. Those third parties are not parties to your contract and are bound to you only through the chain of upstream agreements.
A well-drafted clause requires the provider to impose the same undertakings on its own vendors, which is necessary and settles only part of the problem: it organises liability in cascade without reducing the number of places through which the information passes. The number of links therefore matters as much as their strength, and it is a question of fact.
What is verified instead
The decisive question concerns facts rather than undertakings: what actually leaves the firm, and where does it go? It breaks down into four points, applicable to any provider.
What data leaves the firm's infrastructure, and in what form: the whole document, an extract, a pseudonymised version, a derived representation, metadata?
Where is it processed physically, and under which jurisdiction do those facilities sit?
How long does it remain there, and what is left after processing?
Who, at the provider and at its vendors, can technically access it, and how is that access logged?
A fifth question arises where the provider uses a pseudonymisation process, which is common: where is the correspondence table held, and who holds the key that reads it?
The answer determines the value of the process. Replacing names, companies and addresses with opaque identifiers sharply reduces what a third party could draw from a leak, provided the correspondence is encrypted and the key is not held alongside the encrypted data. If both sit in the same place and are accessible to the same people, the process adds a step without adding a barrier.
A word on vocabulary, which is not a matter of style here. Anonymisation is by definition irreversible; a process permitting identifying elements to be restored is pseudonymisation, and that is necessarily the case in legal work since the final deliverable must carry real names. A provider using the word anonymisation for a reversible process is describing its own system inaccurately, and the imprecision is worth noting.
These five answers describe a state of fact. They can be verified, they depend on nobody's good faith, and they should be obtained in writing and signed by someone who binds the provider.
The contract then regains its full usefulness, on a different object: requiring prior notification of any change affecting the processing or location of data, coupled with a right to terminate. It does not protect secrecy; it protects the firm's knowledge of its own arrangements.
What the client may legitimately ask
One point is worth anticipating, because it now arrives in writing: the client asking its adviser about how its matter is handled.
The request is legitimate and calls for a factual answer. What a firm can usefully say fits in three elements: what processing is carried out, where, and who answers for it. What it should not say is the detail of the tools used, which changes yearly and commits it to a state of fact it does not control.
The request most often comes from a legal department, in the form of a written questionnaire sent to an entire panel of advisers. It is an archived document calling for answers in boxes, not a conversation in which one can qualify.
An answer prepared in the cold, approved by the partnership and kept as a standard response, is worth more than an improvised formulation that will vary from client to client and whose successive versions will eventually be compared.
Secrecy inside the firm
One point is regularly overlooked because it involves no third party: internal access.
A system giving the whole firm a view of every matter creates a secrecy problem inside the firm itself, regardless of any data leaving. The boundaries a firm maintains between teams, for ethical reasons or to prevent conflicts of interest, must be reproduced identically by the system, failing which deployment erases them.
This question arises with equal force in a firm using no external system at all. Internal segregation is an old requirement, and firms have long addressed it by organisational means. A system that ignores those boundaries does not create the problem: it makes it visible, and it aggravates it by granting access to a volume no human organisation allowed anyone to traverse.
This requirement is a technical property rather than a usage instruction: a system unable to distinguish perimeters forces its use to be confined to the least sensitive matters, that is to say those where it helps least.
This article sets out the state of the law at its review date. It does not constitute legal advice and does not replace a professional's analysis of a specific situation.