The deep dive, ten of twelve

The client's protections are sacred.

The declaration most likely to be breached by decent practitioners without anybody noticing. It will not arrive as a scandal; it will arrive as a paste, in the sincere service of the client, in four seconds.

On Declaration X of the manifesto: The client's protections are sacred. By Chris Dias.

The professed seal, pressed in green wax, with the numeral X at its centre

Of the twelve, this is the one I expect to be breached most often, by decent practitioners, without anybody realising it has happened. It will not arrive as a scandal. It will arrive as a paste.

A client's statement into a free consumer chatbot to tidy up the grammar. A witness statement into a summarising tool with an interesting name and a free tier. The other side's without prejudice letter into a general assistant to ask what it thinks. Each of those takes four seconds, each is done in the sincere service of the client, and each one hands a confidence to a third party whose processing terms nobody in the firm has read.

Consider what is being risked in those four seconds. Confidentiality is the foundation, and it is not a courtesy; paragraph 6.3 of the SRA Code requires you to keep the affairs of current and former clients confidential unless disclosure is required or permitted by law or the client consents. Core Duty 6 at the Bar puts it as an unqualified obligation to keep the affairs of each client confidential, with rC15.5 restating it, and the Handbook's own guidance describes that duty as central to the administration of justice. Not central to client service. Central to the administration of justice, which is a much larger claim and the correct one.

Built on top of confidentiality is legal professional privilege, and privilege is the more fragile structure of the two because it depends on the confidence beneath it. Privilege exists so that a person can tell their lawyer the worst thing about their situation and have the whole of it, not the presentable part. Its value is entirely in its reliability; a privilege that might hold is not much use to somebody deciding whether to disclose the conviction they have never mentioned. And a confidence, once dispersed, cannot be gathered back in. The courts have already begun warning that careless use of public AI tools on client matters can put privilege at risk, which produces the sharpest irony in this whole series: casual AI use can destroy the very protection that professing creates.

So the practical discipline is narrower and duller than the principle, and I would rather give the dull version than a stirring one. No client confidence goes into any tool that has not earned it. Earned means somebody in the firm has actually read what the provider does with inputs, whether they are retained, whether they are used for training, where they are processed and who can access them; that there are real undertakings of confidentiality capable of being relied on and enforced, rather than reassuring marketing language on a pricing page; and that the arrangement has been assessed and recorded, not assumed because a colleague uses it. The distinction that matters is not paid versus free, and not cloud versus local. It is whether the confidence stays inside a chain of obligation you can point to.

Two further points that get missed in most firm policies I have seen.

The first is that the danger is the breach itself, and not the consequences of the breach. Practitioners reassure themselves that no harm resulted; the model did not leak, nobody found out, the case settled anyway. That is not the test. The duty is to keep the affairs confidential, and it was broken at the moment of disclosure. Everything after that is luck, and luck is not a compliance strategy.

The second is that pseudonymisation is thinner protection than people think. Stripping the name out of a statement leaves the marriage date, the country, the employer, the refusal history and the medical detail, and in a small community that constellation identifies somebody far more reliably than a name does. I act in immigration matters where a client's circumstances becoming known in the wrong place is not a professional embarrassment; it is a real risk to a real person. I have never once regretted being paranoid about this.

None of which is an argument against using the machine on client work, and I do not want the declaration read that way. I use it constantly. It is an argument for using tools that are fit to receive a confidence, and being willing to accept that the fit ones are less convenient and cost money. The convenience gap between the tool you may use and the tool sitting open in another tab is exactly where this obligation will be lost, and it will be lost by good people in a hurry.

Privilege took centuries to establish and is defended in the higher courts every year by people who understand what it protects. It can be surrendered by one solicitor, on one afternoon, in a single careless prompt.

That is why the word in the declaration is sacred rather than important. It is not the language of risk management. It is the language of something held in trust, on behalf of a person who is not in the room to see how carefully you are holding it.

A protection centuries in the making can be surrendered in a single careless prompt.

Declaration X

Next in the deep dive.

Twelve declarations, taken one at a time, each set against the rules that already bind solicitors and barristers in England and Wales.

Declaration XI: The record proves the vow was kept All twelve