Share It
“Please withhold all attorney-client privileged communication into an isolated review set. Then create a privilege log and read out the document counts.”
This prompt — provided to the generative artificial intelligence (AI) chatbot ChatGPT — demonstrates just one way generative AI tools could increasingly play a role in the eDiscovery profession.
The past year saw an explosion in both the adoption and sophistication of commercially available generative AI tools such as ChatGPT and Stable Diffusion. Microsoft rolled out its new Bing search engine powered by an upgraded version of the same technology that underpins OpenAI’s ChatGPT. Google subsequently launched Bard, its answer to ChatGPT and Bing Chat.
Generative AI has emerged as a legitimate means to produce various types of content, including articles, translations, audio, imagery, and synthetic data, in seconds. These tools can churn out impressively clear responses with exceptional grammar and surprisingly human tone and syntax.
Some experts believe ChatGPT is a tipping point for AI — the culmination of years of technological advances that can transform the very manner in which we write, learn, work, and think. As D. Casey Flaherty, Co-Founder and Chief Strategy Officer at LexFusion, said in a recent webinar on Generative AI, “This is happening. This can be done by you, or this can be done to you.”
What to Watch with Generative AI in eDiscovery
Naturally, generative AI tools have piqued the curiosity of many legal and eDiscovery professionals, eager to learn all of the ways they might be able to help streamline their workflows and save them time and money.
It’s not hard to see why — generative AI tools have numerous potential applications in modern eDiscovery.
For instance, they can be leveraged by an API to drive the instruction set for document review, minus the upfront searching, tagging, organization, and filtering steps.
However, there are a number of challenges to consider. Here’s what to watch with generative AI in eDiscovery:
Slow Adoption
While some legal technology companies may rush to incorporate generative AI into a product to leverage the market buzz, actual adoption of this type of technology has a good chance of being slow. That’s because new technology, in general, typically sees a slower adoption trail in the legal industry due to the industry being risk-averse. This risk aversion is not surprising when you consider lawyers themselves are trained and expected to mitigate risk for their clients. The technology itself is still in its early days and requires considerable effort for it to be safely introduced to the legal profession.
AI has been a part of eDiscovery software for years. In fact, predictive coding or Technology Assisted Review (TAR) has been around for more than a decade. It has evolved quite a bit in that time with active learning, sentiment analysis, natural language processing, and portable predictive models. But even these technologies didn’t see mainstream adoption until just a few years ago.
Not to mention, generative AI is still largely unproven or lacking in precedent within case law in the courts.
In other words, while this technology carries incredible potential, don’t expect it to singlehandedly transform the eDiscovery profession overnight.
Rapid Evolution
Though the adoption of this technology within the legal industry may well be slow and deliberate, the technology underpinning generative AI is evolving at a breakneck speed. These models are advancing in a non-linear fashion, and we are at the front end of an accelerating growth curve. In fact, present-day implementations could potentially seem out of date just a few months from now.
The rate of its evolution will depend largely on the continued investments in its development and training, in addition to the advancement of the broader AI research field. Because of these factors, it’s tough to forecast exactly how quickly these generative AI tools will evolve, but we can expect to see significant improvements in this technology in the coming months and years.
To give a sense of just how fast the AI landscape can move, GPT-4, Open AI’s newest language model, has officially been announced and is already being used in the legal industry. GPT-4 is the first AI tool capable of passing the bar exam.
Unknown Costs
While the prospect of having a live AI assistant is appealing to corporate litigators, the cost implications of leveraging this type of technology are still significant given how new this technology is.
A number of early adopters in the space are already testing possibilities, limitations, and costs — which could provide a glimpse of the feasibility of these tools being deployed in the legal field on a larger scale.
Until we start to get a better understanding of how exactly we’ll be able to utilize these tools more broadly, don’t expect corporate legal teams to replace their existing tools with generative AI.
Legal Considerations
Generative AI tools are already demonstrating remarkable sophistication, but there’s considerable risk in placing too much trust in them too soon.
In other words, generative AI isn’t always right. Just because an AI chatbot’s answers sound polished, they can still blow a basic fact check. These tools are also prone to “hallucinations” and going off the rails.
In the legal profession especially, it’s critical to leverage a human mind to affirm decisions. While these tools may be able to accurately predict a diagnosis based on a patient’s symptoms and medical history, you would still want an actual doctor to confirm it. Similarly, are corporations really going to rely on generative AI when millions of dollars are at stake in major litigation?
Again, as the technology advances, we should expect to see fewer instances of these hallucinations and “fake news.” This is merely an observation of what we’ve seen of generative AI tools thus far.
Data Privacy
OpenAI recently moved to allow users to turn off chat history, giving them control over which conversations can be used to train its models. However, it’s important to consider that generative AI tools acquire their information by systematically scraping the internet. This includes books, websites, comments, and even personal information such as IP addresses, browser types, and user interactions with chatbots. None of this information was paid for or obtained with consent. And, OpenAI states in its privacy policy that it may share users’ personal information with unspecified third parties, without notice, to meet their business objectives.
As is, it’s a potential data privacy nightmare.
Conclusion
It’s hard to argue — AI is changing the eDiscovery profession for the better. At Casepoint, we’ve implored eDiscovery professionals to use AI tools, especially in the earlier stages of the eDiscovery process, for a while now. After all, the amount of data corporate litigators need to corral is growing exponentially in both volume and complexity. This data avalanche is forcing them to embrace new approaches to further narrow the scope of potentially relevant data well before the review stage.
One such approach is Casepoint’s powerful built-in AI and advanced analytics suite, CaseAssist, which helps corporate litigators slash the amount of data that needs to be reviewed; assess each case as early as possible; and cut the time and costs associated with manually reviewing irrelevant documents. In one instance, CaseAssist helped an AmLaw 200 law firm reduce the number of documents that needed to be reviewed by over 90%.
It’s this kind of innovation that will serve legal professionals well over the long haul. That’s why we’re big fans of tools such as ChatGPT. And as the underlying models of generative AI continue to improve, this nascent technology will become more powerful — and potentially useful to eDiscovery professionals.
But though these new generative AI tools have opened up a whole new gamut of possibilities for eDiscovery professionals, the AI technology best equipped to accelerate the most painstaking parts of the eDiscovery process — already existed. So while you should keep a pulse on the generative AI evolution and the ways these AI tools are slowly factoring into eDiscovery professionals’ particular workflows, we think it’s best to exercise some caution. For now, why not utilize the AI that has proven tangible benefits for legal professionals for the better part of a decade?
To learn more about how proven AI technology, such as Casepoint’s AI suite called CaseAssist, can help you overcome your most complex cloud challenges, download our whitepaper now.
