When money, power and AI collide, can lawyers uphold ethics?

Twenty years ago, Bret Walker AO SC asked hard questions about lawyers and money – particularly about how the legal profession’s transformation into “business for profit” threatened traditional ethical values. Today, with the proliferation of AI and leaders internationally threatening the rule of law, those questions have become even more urgent. Can lawyers still defend democracy or have they become part of the problem?

The last two decades have produced some differently inflected outcomes in the ecology of lawyers in the economy. Although the fundamentals are constant, the rule of law and its struggle against error, dishonesty, and their motivating bases of greed and power, is a composite of very considerable human force.

When unbridled political power clashes with the professional duties of ethical lawyers, who are trying to make a good living in the markets for their services – bad things happen that good people should resist. However, this is much more easily said than done.

The Trump administration’s abuse of its capacity – to wield power as a customer, as a prosecutor, and as an informal regulator of legal practice – is deplorable. Clients must be able to hire and fire their chosen lawyers at will, for good or bad reasons. We have no right to be retained. Mind you, a government’s wilful, and ultimately wasteful counterproductive firing of good lawyers, and hiring of charlatans in their place ought to arouse an outcry against its corrupt diversion of public money. Abuse of the power to investigate suspected crime and prosecute it is a notorious habit of authoritarians throughout history and around the globe. It’s a shame to be witnessing it in a supposed constitutional democracy. And spurious rules imposed by social media and political figures that restrict free speech or policy choices – especially those related to diversity, equity and inclusion (DEI) – are being used to punish big law firms. That all adds to the oppression. As does the campaign to punish lawyers for their temerity in having advised or represented the president’s foes, real or imagined.

The ‘money imperative’ lying on big firms explains the scandalous extremes of the Trump administration’s malign influence on these prominent lawyers in the United States. For example, the current resolution of the tensions in question requires these big firms to donate – as if by way of protection money, which are mislabelled ‘pro bono services to the administration’ to be deployed at the command of the President and his henchman in the White House. So, how does this mess implicate lawyers and money? Because it appears that but for the financial disaster intentionally threatened by the Trump’s attack on the big firms, many, and maybe most of them would have repudiated the attempt and refused to grovel. Enough partners protested and resigned, and enough young lawyers were appalled, and also resigned, to indicate there still values of honour and professional pride that can produce the proper stream of resistance to the political abuse of lawyers.

Principles or survival presents a choice none of us can lightly confront. We should feel great sympathy for the leaders of big firms who must have known that knuckling under would see them branded as cowards. And that defiance might require life altering career changes. This dilemma is massively worsened by the sheer scale of big firm operations. Whatever the advantages of big firm operations it is in this case, a Faustian bargain.

Now I want to turn to the provocative letters ‘AI’ and its impact on lawyers, clients and the administration of justice. If ‘A’ is the opposite of real, they could be read as a plain confession of inferiority to human intelligence. If ‘I’ is intelligence in the sense of knowledge as in the expression ‘criminal intelligence’, which refers to information available to police, not the brain power of criminals, then we are well aware of the inherent weakness of intelligence in that sense. Because it is far closer in resemblance to mere belief or suspicion than to actual, let alone, demonstrable, knowledge.

One thing is clear. The words ‘wisdom’ and its subsidiary concept ‘judgement’ are not at all entailed in AI. History suggests we should not disdain the fact that an intellectual resource is artificial. To the contrary, reading and writing are at the pinnacle of our species artificial achievements, as well as tools. So is law, which both predates and flourishes with literacy. So is government, as shown by its myriad forms and its constant plasticity. In ordinary legal practice, indices, digests, and treatises are all artificial aids to professional intelligence, in both main senses of the word. They do help us to think, to argue and to resolve the problems our clients bring to us.

My comments are not directed to the innocent use, entirely beneficial, of digital resources for collating and retrieving the mass of written materials that are the stuff of legal practice. On the other hand, the use of so-called precedents for written legal work – whether they’re contracts or pleadings, opinions or advice on evidence, or reasons for judgement– is well understood to have inherent hazards. By their nature, precedents either oversimplify or overcomplicate. They are not custom made. They are difficult to adapt to the individual and particular facts and circumstances of a matter requiring legal services being a matter unknown before the precedent was devised. And precedents hinder innovation designedly.

AI as a generator of fully expressed legal analysis and advice is simply wrong. No financial dimension or consequence alleviates that condition. AI perverts its useful purposes. The profession doesn’t need a digital prompt to prepare advice which identifies the issues, the client’s aim, the material facts, the applicable law, the attendant risks, or the possible outcomes and prospects. Unless the profession of course is lazy, or worse, has an inclination to pretence.

I have no time for the idiotic notion of AI-generated submissions or evidence such as affidavits in the courts. But with submissions, we must remember how we make them – not only by the formation of judgement, but by the obtaining of knowledge of relevant facts and law. They are conscious of the client’s individual circumstances and vulnerabilities. And above all, we make them in compliance with our ethical duties of candour and participation in the administration of justice. Which is not a sporting contest, let alone an electronic game. When it comes to the scandal of ‘AI hallucinations’ – which in my view, is too kind a label – that invent phoney cases as authorities and fictitious statutory provisions, I’m sure the coders and others, who for financial return are responsible for the algorithms that produce such garbage, surely did not specifically instruct the processes to those ends. But how do these features of AI exist and persist in this aspect of legal practice? We can probably blame large language models (LLM). Which is not a further degree obtained from a university, but models that trawl existing publications and conduct pattern recognition for loosely described predictive results. But patterns are a construct of observation, and they are constantly changing in hard science. And predictions are always a mixture of chance and falsity.

AI can’t stand in for ethical legal work any more than pre-digital resources such as indices and treatises. No amount of financial incentive of savings, enhanced revenue or increased profit margin can say anything to the contrary of that proposition.

So what is the legal profession to do? Do we supplant, supplement, or sharpen the tools that we presently have, over and above the latest Court commandments which have wielded an already salutary effect in the profession, against the absurdly overhyped use of AI in the administration of justice? I have a modest proposal that invokes a modest slogan of user pays. In this case, the misuser pays. Apart from being struck off, any lawyer who misleads a court, opponent or client, by misusing AI in place of their own research, thought and articulation of the result of those processes, should also play full indemnity costs personally for all the time and trouble wasted by that delinquency, and of course, partnership and vicarious liability as appropriate within firms. This approach is easily imaginable. It’s quite feasible in the realm of litigation where the peremptory and condign visiting of costs consequences on misbehaving lawyers is a healthy and old tradition. It’s more difficult in non-contentious work as there aren’t clear ways to recover costs. However current statutory disciplinary proceedings may accommodate some compensation. But it’s time, for the profession to devise and to propose to government a concerted approach to ensure that there is no continuing financial benefit for lawyers misusing AI in any kind of work. This idea opens the possibility that a non-delegable duty might arise from ordinary common law duties of care which would treat the promulgation of AI with its inherent feature of hallucinations – truly to be known as lies – as just another form of ‘snails in the bottle’.

After all, resorting to LLM is a resort to the mass-assembled, collated and analysed verbiage of humans. Humans, being the species of animal which lies. That propensity is known. AI’s risks are obviously foreseeable. It could be described in terms recognisable from Rylands v Fletcher – the territory is obviously inherently dangerous. The liability should follow in near absolute terms. But that contention, which seeks to embrace AI and money and not merely lawyers and money, is one which is beyond my remit.

This is an abridged version of Bret Walker AO SC’s speech, Lawyers and Money, 20 Years Later.