Hello Everyone,
Summer is approaching and so far 2024 is shaping up to be a pivotal year for Generative AI’s impact on the legal profession. I’m half expecting Microsoft to develop a Copilot for the legal industry.
This is a guest post by one of our most prolific guest contributors. As a lawyer and on the ethical side, Tobias explores the AI landscape in a valuable lens.
Recent work by Futuristic Lawyer
AI Music Is Not Art. It’s… The Opposite
How Will AI Change Law & the Legal Industry? (Level 1)
Understand the Essentials of the Digital Markets Act & the Digital Services Acts
To Open-Source or Not to Open-Source?
Tobias regularly appears on Podcasts if you would like to reach out to him you can do so here.
Key Partnerships in the Future of Legal AI are Forming
In recent times Harvey AI backed by Sam Altman (OpenAI Startup Fund) has partnered with Mistral, while earlier this year Robin AI partnered with Anthorpic. If I was Microsoft, I’d also build a Copilot for Lawyers and the legal profession. In my view, few professions stand to gain as much from Generative AI and specialized Copilots as law and the legal profession as a whole.
Like software development and coding, Law stands to gain so much time from the legal Copilots of the future.
👨🏻🏫 I’m trying to bring in more guest contributors who are experts in their field for the benefit of my readers. If you think this warrants support of just $2 a week, it will also enable you to access more deep dives, feel free to support.
This publication’s future depends on paid subscriptions for the bulk of its revenue that is my primary income source. If you enjoy what I am doing with the sum-total of my Newsletters on emerging tech. Follow me on Notes too.
GitHub Copilot but for Lawyers?
The market for legal copilots has exploded in 2023-2024. Hundreds of tech companies are lining up and competing to supercharge English-speaking law firms with generative AI capabilities for drafting and reviewing contracts, providing legal chatbot support, analyzing case law, summarizing and searching through long documents and databases at the speed and proficiency of ten thousand junior lawyers. The envisaged end-destination is presumably a full-on robolawyer who can be hired as a new colleague, much like Devin, the autonomous AI software engineer.
Several parallels exist between coding and contracting. Both disciplines are rule-based and logical, follow common standards, norms, and predictable patterns, and are used to describe and achieve specific, desired outcomes. Considering how transformative AI has been for coders, it’s not surprising that a lot of companies are trying to make an equal impact in the practice of law.
For example, the CEO and founder of Spellbook, an AI legal startup that brands itself as “ChatGPT for Lawyers”, Scott Stevenson, explains in an interview how his experience with GitHub Copilot inspired him to pursue a corresponding tool for lawyers (via Bret Kinsella from Synthedia).
“The first time I used it, it outputs 20 lines of code that I thought were garbage and I thought, ‘Wow, this doesn’t work. This tech isn’t quite there yet.’ And then 10 minutes later I realized that the code was anticipating an issue that I hadn’t thought about myself.
That was the first time I felt outsmarted by AI. It was truly thinking 10 minutes ahead of me and had discovered a problem I didn’t see and it had written more robust code than what I’d been thinking of to deal with and mitigate that problem. That moment was when we decided, ‘This has to exist for lawyers. This had to exist for contracts.”
GitHub Copilot launched for technical preview in June 2021. It started as an autocompletion tool that suggests lines, comments, and functions to grinding coders.
An early study indicated that Copilot made coders work 55% faster. An early survey showed between 60–75% of Copilot users reported feeling more fulfilled with their job, less frustrated when coding, and able to focus on more satisfying work. 73% said Copilot helped them to stay in the flow, and 87% said it helped them to preserve mental effort during repetitive tasks.
Since then, Copilot has evolved quite a lot. It’s integrated with GPT-4 with a new chat and voice interface, and it has the ability to recognize and explain code, recommend changes, and fix bugs. The latest development is GitHub Copilot Workspace which enables coders – and non-coders – to “brainstorm, plan, build, test, and run code in natural language”. GitHub writes in the announcement post:
“Early last year, GitHub celebrated over 100 million developers on our platform—and counting. As programming in natural language lowers the barrier of entry to who can build software, we are accelerating to a near future where one billion people on GitHub will control a machine just as easily as they ride a bicycle.”
The pace of development is staggering. Three years ago, Copilot was a simple autocomplete tool. Now it’s more like a genie in a bottle that can fulfill coding requests like an autonomous software engineer. We also have to emphasize here with a fat marker that Copilot’s staggering pace of development is not without issues or concerns.
GitHub, Microsoft, and OpenAI were defendants in the first among a series of class-action copyright lawsuits filed by Matthew Butterick and Joseph Saveri Law Firm. The lawsuit from October 2022, is aimed at how GitHub Copilot was trained on GitHub’s code repository without rightful attribution to the original creators. Besides the copyright question, other concerns are code quality and overreliance. A group of researchers from NYU – Hammond Pearce, Baleegh Ahmad, Brendan Dolan-Gavitt, Ramesh Karri, and Benjamin Tan from the University of Calgary – found in a study from December 2021 that approximately 40% of Copilot’s code contributions were vulnerable. See also this YouTube review of the Copilot Workspace announcement by software dev Theo – t3․gg.
At the same time, working with AI tools is now standard practice for developers. A survey by GitHub from last year showed that 92% of 500 surveyed developers admitted to using AI coding tools at work and for outside projects. AI copilots are also changing how coding is taught.
Can GitHub’s success with Copilot be replicated with a similar tool for lawyers?
🏅 Featured Article
By of LawVerse
An exploration of AI and copyright.
Microsoft Copilot but for Lawyers?
There is another copilot we need to talk about.
(A cringeworthy promo video for Microsoft Copilot that looks like it should have been a Nike commercial)
Microsoft rebranded Bing Chat, the early version of GPT-4 with internet access, to Microsoft Copilot in February 2023. Today, Microsoft Copilot is an “everyday AI companion” with over 150 features. Microsoft’s Copilot can be used as a writing assistant in Word or Outlook, an editing tool in Paint or Clipchamp, a shopping helper in Microsoft Shopping, take notes during Team calls, generate images with DALL-E 3, songs with Suno, and much more. For the cost of $30/month per Business or Enterprise user and $20/month per Personal or Family user, Microsoft’s suite of products is sprinkled with magical AI fairy dust.
The Copilot idea has become a major bet and a core part of Microsoft’s vision for the future of their business. Not only did Microsoft make the first change to its keyboard in 30 years, by adding a new Copilot key, but a whole new Copilot Plus PC is coming out later this year.
Microsoft 365 is a common digital workspace for many, many law firms, so integrating a smooth-running legal copilot with the 365-package, “a Clippy on steroids for lawyers”, is key to leveraging generative AI in legal workflows. Below are a few suggestions on how to use Copilot for Microsoft 365 from a 16-page Microsoft publication titled Generative AI for Lawyers (2024).
Many legal departments around the world are already using Microsoft Copilot as outlined above. Companies are also taking advantage of Microsoft Azure’s enterprise data security and privacy safeguards, and building their own copilots and generative AI applications on Azure OpenAI Service. For example:
The well-established legal tech company, Litera, announced on May 1 that it would deepen its partnership with Microsoft to bring new tools to market and streamline workflows for lawyers with Copilot for Microsoft 365. Litera’s, Head of Product, Adam Ryan, said: “We believe that by the end of 2024, many lawyers will be using Copilot in their day-to-day.” Litera also owns Kira Systems, “a patented machine learning software that identifies, extracts, and analyzes content in your contracts” that is used by many law firms.
Legal professionals at KPMG have used Microsoft Copilot since last year and are partnering with Microsoft to prepare clients “to adopt and implement generative AI within their legal teams and helping those teams advise on how to navigate the shifting legal and regulatory landscape”.
Thomson Reuters announced last year that it would invest more than $100 million annually in AI capabilities and add a new plug-in to Microsoft 365 Copilot. According to the press release: “This integration will bolster efforts for redefined professional work starting with legal research, drafting, and client collaboration.”
None of these partnerships are likely to make legal copilots a common household item. Rather, the partnerships are geared toward the companies’ own internal use of Copilots’ technology and
implementing it into existing product offerings.
Chances are high that the golden standard for a stand-alone legal copilot will not be created by Microsoft but by market forces. Hundreds, if not thousands, of promising candidates, are currently competing to make the next GitHub Copilot but for lawyers. Who knows, once the dust has settled one of the major winners may be acquired by Microsoft. Such are the rules of the coopting disruption playbook, the ruleset BigTech plays by.
Legal Copilots as of May, 2024
There are many promising legal AI startups and related products in development in 2024, such as:
Casetext
Describing the market for (potential) legal copilots and doing it justice in a single newsletter post is a tough order. There are too many to choose from. Yet, I want to give you a brief overview of the market’s top segment based on funding, size, and customer base.
The London-based Robin AI raised $26 million in Series B funding at the beginning of this year. The company is one of Anthropic’s launch partners and the only dedicated generative AI tool for lawyers that is based on Anthropic’s foundation model, Claude. Robin AI’s legal copilot is available as a Microsoft Word add-in, and can allegedly cut the time it takes to review contracts by 85% (Tech EU).
Spellbook is an AI legal startup that sells “ChatGPT for lawyers” and “uses GPT-4 to review and suggest language for your contracts” in Microsoft Word. It announced on January 24, 2024, that it had raised $20 million in Series A funding. With this new funding, it plans to scale rapidly into 30,000 law firms worldwide (from 1.700 law firms at the time of the announcement), while working towards developing a full end-to-end AI workflow that mimics the style of a lawyer.
Draftwise offers an AI solution that empowers transactional lawyers from contract drafting to negotiations. The company raised $20 million in March and was founded by two former Palantir engineers and a lawyer at the global law firm Clifford Chance (Reuters).
Then we have Harvey, the mysterious Legal AI startup that raised $80 million in Series B funding in December 2023, at a $715 million valuation, without releasing as much as a product description or a demo. Harvey was one of the first four startup investments by the OpenAI Startup Fund (at the time controlled by Sam Altman) and it raised $21 million in a Series A funding round led by Sequoia earlier in 2023.
Just recently, the company published a product page on its website that gives a peak into Harvey’s functionality. Based on this page, Harvey can help to draft and analyze any type of legal document, answer complex legal questions across multiple domains with citations linking to the source material, identify issues in a contract, orchestrate workflows across hundreds of LLMs, provide key insights into a firm’s productivity, and more.
Harvey is in stiff competition with another San Francisco-based and 9-year-old company, Casetext, which offers an AI legal assistant called “CoCounsel”. Casetext serves more than 10,000 customers from small in-house teams and solo practitioners to big law firms such as DLA Piper and US-based Fisher Phillips LLP. Last June, Casetext was acquired by Thomas Reuters for $650 million in cash.
London-based, Luminance, is another major Harvey competitor. It is currently expanding its services to the US on the back of a $40 million Series B funding round. In November of last year, Luminance showcased the first completely AI-powered contract negotiation. According to its own testimony, Luminance services all of the Big Four consultancy firms, over one-quarter of The Global Top 100 law firms, and multinational organizations such as Koch Industries, Hitachi, and Avianca Airlines.
ContractPodAi is another London-based company and it was founded way back in 2012. It delivers contract management software as well as a customizable legal copilot called “Leah” that is used to draft and redline contracts, analyze critical patterns and insights, extract data in legal documents, answer legal questions, write playbooks and guidelines, and bring AI support to contract negotiations. ContractPodAI raised $115 million in growth funding in its latest series C funding round in September 2021.
Superlegal is an offshoot of Lawgeex, an Israeli-based, “old-timer” founded in 2014. Lawgeex recently had to sell off its clients and assets to RobinAI and another AI contract management platform called LegalSifter. Lawgeex CEO and co-founder, Noory Bechor, is now leading Superlegal and the company raised $5 million in seed funding just last Thursday.
Superlegal claims that its AI reviews contracts better, faster, and at a lower cost than any lawyer. The company aims to provide a cost reduction in legal fees of 90% to small and medium businesses. It reviews customers’ contracts within 24 hours and also has human professional lawyers in the loop – the same model was applied by Lawgeex. The only difference is that Superlegal aims at supporting medium-sized companies with full-automated legal help, whereas Lawgeex’s focus was giant corporations in the Fortune 500 and Global 2000 category. Superlegal has also kept hold of Lawgeex’s patented technology and its license to practice law in Utah.
Wrapping Up
Currently, new competitors and solutions that promise to revolutionize and transform law are emerging on a near-weekly basis. However, a company that is developing a legal copilot has to answer a few tough questions that are unique to the legal industry. Different from GitHub Copilot or Copilot for Microsoft 365, makers of a legal copilot are forced to consider the rather sacred attorney-client privilege, liability for wrongful advice, lawyer insurance, and more. Privacy and security fears could in my opinion lead many big law firms and companies to develop legal copilots in-house. I would expect the market for legal copilot to consolidate over the next few years but let’s check in again in 6-12 months.
More by the Author
Futuristic Lawyer is becoming one of my go-to sources for AI policy, ethics, law and many perspectives also from a more EU centric view. Tobias hails from Copenhagen, Denmark.
The work of Tobias Jensen always feels grounded in a kind of common sense, research and EU perspective that might be impacted by his background in Law. I find it resonates with my own cautious view towards pro-AI technological optimism that doesn’t always take the legal, security, public interest and accountability aspects into full consideration. I personally learn more from these kinds of writers than I do more popular AI evangelists.
How to Deal with Data Harvesting AI Girlfriends?
AI’s Labor Market Impacts Are Slower Than Expected
Thoughts on AI Writing Assistance
A Sensible Approach to Regulating AI in Copyright Law – The Input Phase
Read More in AI Supremacy
