Why lawyers' guiding principles are still relevant in the age of generative AI

In this blog, Eigen's Head of Legal, Alice Sahba, explores why the timeless guiding principles of ownership, value protection and downside mitigation are still relevant today.

I have worked as a lawyer at organizations ranging from the world’s largest provider of financial markets data to a global law firm, an alternate finance provider and a disruptive start up, selling products and services to the most diverse client base imaginable. Yet the sticking points at the contractual stage of B2B agreements — ownership, value protection and downside mitigation — are reliably the same at all of these organizations, regardless of the banner under which they are presented.

In my experience, no matter the industry, size or geography, the struggle of sellers and buyers follows a well-trodden path:

1. Ownership

Seller: Seller wants to sell on an “as is” basis, disclaim as much liability as possible and Buyer buys at its own risk.

Buyer: Buyer wants assurance as to what it is buying, that it will have full rights to what it is purchasing and rights of recourse if any of it turns out to be untrue.

2. Value Protection

Seller: Seller wants to protect the value of what it is selling to avoid weakening its position in the market. Seller does not want any restrictions on its right to sell, while applying restrictions on what Buyer can do with what it has purchased.

Buyer: Buyer wants to ensure the confidentiality of what it is sharing with Seller, to similarly avoid weakening its position in the market. Buyer wants to restrict Seller’s right to reuse any insights gained from Buyer while ensuring the widest rights of use of what it has itself purchased from Seller.

3. Downside Mitigation

Seller: Seller wants to disclaim as much liability as possible and does not want to assume any amount of liability above the value of the contract.

Buyer: Buyer wants there to be recourse if it suffers losses greater than the value of the contract because of something the Seller has done or has failed to do. Buyer could be exposed to lose a lot more than it would have ever been set to gain.

Depending on the respective bargaining powers of Seller and Buyer, and their lawyers’ penchant for point scoring, this struggle ends somewhere in the middle of these two positions. Many a lawyer has missed out on important family events, dinner parties or weekend plans in the name of being able to find a mutually acceptable compromise. A sometimes-tedious process (that technology will soon also disrupt), but nevertheless an understood necessary evil to maximise business objectives while mitigating risks, and ultimately protect long term value for businesses and their stakeholders.

And then, in November 2022, Chat-GPT kicked off an AI arms race that resulted in what can only be described as an enormous dose of business FOMO. The mandate was clear: find ways to leverage the power of these LLMs or risk being left behind. The name of the game: speed.

The legal community was equally quick to report on the risks associated with using platforms such as GPT. However, companies (those without integrated legal support and, worryingly, even those with it) still seem intent on throwing themselves into this pursuit without regard to their three guiding principles of Ownership, Value Protection & Downside Mitigation and, crucially, without regard for whether these LLMs and the risks associated with them are the best solutions to meet their business goals.

To briefly restate, here are the terms that businesses accept when using the GPT family of tools based on OpenAI’s current terms of use and the existing legal landscape:

A) Ownership: User does not own the generated output; it is unclear who owns the output, if it is even capable of being owned. On top of that, although the User is being given rights in the generated output, given the way that the LLMs are trained, there is a good chance that the output has been generated in a way that infringes the intellectual property rights of many third parties. User is in essence not receiving ‘clean’ rights in the output and, if incorporated into business processes, those outputs risk tainting other business assets. 

B) Value Protection: Users share valuable information when they interact with the LLMs and, although there are ways to limit OpenAI’s further use of that data, it is still unclear what rights are granted to OpenAI in the input/user data and there is, compared with the industry standards for value protection, a weak protection of the privacy and confidentiality of the input data.

C) Downside Mitigation: User not only has no recourse to OpenAI in the event that it suffers loss as a result of using the generated output but is itself fully on the hook to OpenAI in the event that OpenAI’s service has yielded results that infringe the rights of third parties. Liability is unlimited and uncapped.  

I am the first to admit that over-lawyering is a real problem that can stifle innovation, however wilful ignorance about the risks of using LLMs in their current form, and in the current landscape, whether being risks focussed on the principles of Ownership, Value Protection & Downside Mitigation, or other risks, of which there are many, is not protecting long term value for business stakeholders. The power and potential uses of LLMs are mind-boggling, but we should not lose sight of these risks and we should not be afraid to challenge whether they are in fact the best solutions to meet all our business goals.

I have no doubt that the landscape and lens from which we will look at these questions will change but, for the time being, these principles (centuries in the making) should not be quickly disregarded. Generative AI will change the way we all work, but it is not necessarily the best answer to every problem. Anecdotally, I know that many institutions are now turning their noses up at thoughtfully developed and proven technologies in the hopes that GPT can solve their every business need.

In the context of Intelligent Document Processing (IDP), Eigen Technologies has spent the last 8 years leveraging, developing and fine-tuning artificial intelligence to extract data from documents at scale. In contrast to OpenAI’s offering, Eigen has thoughtfully developed its solution in such a way that it can offer its customers what they are looking for, namely: speed, accuracy, transparency and cost-efficiency, as well as a customer centric position on the guiding principles of Ownership, Value Protection & Downside Mitigation.

To find out more about Eigen’s IDP solution, visit our Platform page or set up a short call with one of our automation specialists here.