Big AI v Big Law — Jude Law, Harvey Specter and why the legal AI arms race is only just beginning

Alex Heshmaty considers the impact of the latest developments in the legal AI landscape.

Since ChatGPT was released towards the tail end of 2022, many of those working in professional services industries have been monitoring the huge strides made in generative artificial intelligence (AI) with a mixture of excitement, bemusement and some degree of existential dread.

Although the initial impact of AI was felt by copywriters, translators and graphic designers – many of whom lost their livelihoods as a result – 2026 is shaping up to be the turn of the legal sector to grapple with the full force of the growing waves of AI.

Big AI and Big Law

One of the most iconic images so far of Big AI reaching out to Big Law was the recent marketing campaign by Legora featuring Hollywood star Jude Law. Rival Harvey opted to make its namesake from TV drama Suits its brand ambassador. No doubt the marketing budgets wielded to secure A-list celebrities must be eye watering, commensurate with the company valuations of $5.6 billion and $11 billion, respectively.

But these costly marketing efforts are unlikely to make a dent in a potential rival valued at almost $1 trillion: Anthropic. The frontier large language model (LLM) created waves across the legal sector when it unveiled a legal plugin for its Claude product, immediately hitting the stock price of incumbent legal publishing behemoths such as LexisNexis and Westlaw, which have their own legal AI offerings. (‘Frontier LLMs’ denote the most advanced AI platforms available – typically the latest models of ‘foundational LLMs’.) Microsoft, which owns a 27% stake in OpenAI, quickly followed suit by launching a legal AI agent in Word.

Meanwhile, several companies – spanning law, legal tech and legal publishing – are attempting to develop their own foundational AI platforms. (‘Foundational LLMs’, sometimes called ‘core LLMs’, are the proprietary AI platforms trained on vast amounts of raw data – as opposed to ‘wrappers’, which provide an AI interface based on a third-party foundational LLM.) US firm Kirkland & Ellis has set aside $500 million to create its own AI tool, which aims to hoover up the “collective intelligence” of all its partners. Meanwhile both Harvey and Thomson Reuters are attempting to create their own LLMs from the ground up.

Big AI: threats and opportunities for the legal sector

The fast-moving developments in the legal AI space are making it challenging for law firms to know what parts of AI to embrace while navigating the risks. Brian Inkster, founder and CEO of law firm Inksters, says that despite the obvious disruption posed by legal AI, it does not yet pose an “existential threat” to the legal profession. However, it’s “a minefield at the moment and an ever-changing one” for law firms, and he urges them to consider what problem (if any) they are trying to solve with AI, rather than simply rushing to adopt it for fear of missing out.

He warns that “hallucinations are still a big problem and one that is not going away anytime soon”, emphasising the high burden of verification necessary if using AI for legal research and noting that crucial information can be missed. Of the incumbent legal publishers who have adopted AI, he says that “LexisNexis and Westlaw are arguably less reliable with an AI layer added than they were before it was added”.

But the potential for AI to improve access to justice cannot be ignored, according to Joanne Brook, consultant at multi-disciplinary firm Brecher: “Tools that help lawyers will always be welcomed – even ones that replace lawyers – if they provide broader access to legal services.” Nevertheless, she is concerned that the cost of implementing AI could mean that it will “only be adopted by medium to large law firms who have to charge more to support the licensing and legal technicians”, thereby pricing out access to legal advice for many potential clients.

Is legal AI eating itself?

Brian explains that the legal AI-specific vendors (such as Harvey and Legora) are, so far, “just a wrapper over the core LLMs. They rely on those core LLMs. As do the incumbent legal publishing giants when they add an AI layer.” As such, he believes that, as foundational AI companies decide to verticalise and increasingly offer products with a legal focus, many of these ‘wrappers’ could fall by the wayside; law firms could just “eat the banana and throw away the skin”. He believes that the aforementioned legal agent rolled out by Microsoft could prove particularly disruptive, as “Word is where most lawyers live and an AI legal-specific agent embedded there may be all lawyers really need”.

Jo concurs that the threat posed by foundational LLMs to legal AI-specific vendors is very real: “Legal work is an obvious target for core LLMs with precedent, codified law, technical language and clear workflows providing an early training boost upstream and downstream. Will smaller vendors be snapped up? Of course! The big fish vendors need to ensure the waters are clear of competitors.”

Will legal AI put firms on a level playing field or widen the gap?

The gap between the largest firms and their smaller rivals is ever widening according to Jo, who warns that if “large law firms pull away in terms of adoption of new technology, scale and pace”, there might be no way for smaller firms to catch up, other than through focusing on specialisation. In this scenario, she believes the cost of legal services could rise overall, leading to a reduction in access to legal advice.

Brian offers a glimmer of hope for smaller firms hoping to adopt AI, saying that legal AI vendors are trying to appeal to a wider section of the legal market beyond the Magic Circle: “It is notable that when Harvey launched, and for some time thereafter, they remained the preserve of Big Law firms only. They were not interested in speaking to smaller law firms. Now they appear to be open to anyone. I assume with all the investment that they have received they now need to increase sales. Thus, they are having to spread themselves thinner.”

Brian also notes that open source legal AI products have started appearing: “Mike OSS claims to give you an open source platform to rival Harvey and Legora. So smaller firms with a bit of technical know-how could perhaps do their own thing at a fraction of the cost.” But he believes that Microsoft’s legal agent in Word could potentially “level the playing field like nothing else might” in time.

What does the future hold for the legal profession in the wake of AI?

In the short term, the importance of AI training for lawyers will increase according to Jo: “Not only do new lawyers need to know how to properly use AI with appropriate search prompts and interpretation and selection tools, they need to know how to interrogate the output.”

In terms of AI policy, she urges the legal profession to “hold the tech companies to the same standard as it holds legal colleagues to ensure that AI legal tools meet – or can only be used in a way that meets – the core obligations of legal practice by human lawyers, as set out in the Law Society of Scotland Standards of Conduct”. This includes issues such as maintaining client confidentiality and bias contained within training data of the LLM models. She also says that lawyers need to be “thinking much more about [the] insurance risk of using AI” and, more broadly, the profession as a whole needs to be having a debate about “what it wants to remain and what it wants to become”.

According to Brian, in the short to medium term, law firms that have “rushed in head-first without much thought, guardrails, checks and balances” could face a “very messy future with a lot of cleaning up required”. This is perhaps evident in the increasing number of high-profile cases of law firms being caught out for not removing AI hallucinations from court submissions.

In the longer term, things are “likely to settle down a bit and it will become clearer where actual benefits lie” – which could prove advantageous to those who “sat back and waited” rather than hastily adopting AI. But Brian says that the cost of AI licensing will be the “thing to watch” as the AI companies are currently “heavily subsidised and the core LLMs are losing money by the truck load. Once you are hooked in and the prices skyrocket, will it be sustainable?” The recently announced introduction of consumption-based pricing by Legora – which will almost certainly lead to higher costs for law firms – is perhaps a sign of things to come.

Considering the speed of developments in the legal AI space, there could be some significant changes over the next couple of years. The interplay between legal AI-specific vendors, incumbent legal publishers and foundational LLMs will likely see a few winners but probably many more losers. The war chests of Anthropic, OpenAI/Microsoft, Google, X and Meta – and the fact that currently all the legal AI-specific vendors rely on their models – will likely lead to them acquiring the bulk of their competitors.

Although some of the biggest law firms might initially succeed in building their own AI models, ultimately the cost of keeping up with the frontier LLMs will probably lead to most firms relying on the handful of AI behemoths. What is perhaps more uncertain is whether large portions of the legal industry itself could be swallowed up by the whales of Silicon Valley.