Guest post by John Hayes, managing partner of London employment and regulatory firm Constantine Law
Hayes: Is an AI-generated internal summary of legal advice privileged?
Much has been written about the risks of lawyers misusing AI for legal research and drafting. Recent cases involving fabricated authorities and inaccurate legal analysis have rightly attracted attention.
However, in my view, the greater and more immediate challenge lies elsewhere: the routine use of AI by clients themselves.
Every day, clients are uploading legal advice, correspondence and draft documents into AI platforms such as ChatGPT, Claude and others.
They do so for entirely understandable reasons – to critique legal advice, test its accuracy, summarise it for colleagues, or make it more accessible to decision-makers within their organisations. Yet in doing so, they may be creating significant risks for themselves.
Consider a recent example from our firm. A client responded to a draft letter prepared by one of our partners with the observation: “Congratulations – you got 19 out of 20.”
When asked what he meant, the client explained that he had uploaded the letter into an AI platform, which had scored it 19 out of 20 for accuracy.
What followed was a fascinating discussion. Our partner was ultimately able to demonstrate that the AI model had failed to appreciate a critical factual nuance addressed in the letter. In other words, the missing point was not an error by the lawyer but by the AI system itself.
At one level, this is an amusing story. At another, it illustrates a profound shift. For the first time, lawyers can expect their work to be routinely reviewed, scored and challenged by clients armed with increasingly sophisticated AI tools.
The democratisation of legal knowledge is already under way. AI has given non-lawyers unprecedented access to legal analysis and reasoning, enabling many to act as relatively effective first-instance advisers to themselves.
The consequences are already visible. As an employment law practice, we now routinely receive draft grievances and complaints generated with the assistance of AI that run to 15 or 20 pages. A few years ago, equivalent documents were often four or five pages long.
Ironically, AI is not currently reducing the volume of legal correspondence; it is increasing it.
However, the expansion of legal paperwork is not the principal concern.
The more serious issue is the effect that clients’ use of AI may have on confidentiality and legal privilege.
We are increasingly seeing clients upload legal advice into AI systems for the purpose of generating internal summaries.
In one recent matter, a client uploaded a detailed letter from Constantine Law into Claude and asked the system to produce a short summary that could be forwarded to a line manager. The summary was then circulated internally.
That seemingly innocuous act raises two important concerns.
First, confidential and legally privileged communications may be disclosed to a third-party AI platform. The legal consequences will depend on the circumstances, the platform used, its terms of service and the applicable law.
However, clients should not assume that uploading privileged material into an AI system is risk-free. The possibility of waiving confidentiality or privilege must be taken seriously.
Secondly, once an AI-generated summary is created and circulated, questions arise as to whether that summary itself attracts legal privilege. The summary is no longer the lawyer’s communication. It is a new document produced through the interaction of the user and the AI system. Whether privilege attaches to that document may be far from straightforward.
In highly sensitive and contested disputes – which describes many employment cases – these distinctions matter enormously. If privilege is lost or successfully challenged, communications that clients assumed were protected may become disclosable in litigation.
Internal exchanges between managers discussing AI-generated summaries of legal advice could become fertile ground for disclosure requests.
The danger is that clients may unwittingly undermine one of the most valuable protections the legal system affords them.
There is no easy solution. The reality is that AI is now embedded in the workplace. Clients will continue to use these tools to analyse, challenge and summarise legal advice. That genie is not going back into the bottle.
The law of privilege has traditionally been interpreted narrowly and is easily compromised by careless dissemination of legal advice. Whether the law will evolve to accommodate the realities of AI-assisted communication remains an open question.
It may be that courts and legislators will ultimately develop new principles that reflect modern business practices.
For now, however, the law has not fully caught up with the practical realities of AI.
Clients therefore need to understand that the unrestricted use of AI tools in relation to legal advice carries real risks. In some circumstances, it may jeopardise confidentiality, weaken claims to privilege, and erode one of the principal benefits of seeking legal advice in the first place.
Much remains to be debated and resolved. Yet from where I sit, the most significant AI challenge facing the legal profession is not the use of AI by lawyers. It is the use of AI by clients.
Leave a Comment