Experts discuss the evolution of privilege review as technologies improve
A thorough and accurate privilege review is as important today as it was decades ago, maybe more so given the amount of sensitive information organizations store. But the industry has reached a point where eyes-on reviews are cumbersome and riddled with risks and mistakes. Massive volumes of data make manual reviews impracticable.
What can the industry do other than evolve? That was the topic of the latest webinar hosted by H5 and Corporate Counsel Business Journal.
The panelists began by looking at current privilege review methodologies, and the risks firms and organizations have to address. Current workflows rely on multiple levels of manual review. The level one reviewers code for relevance and privilege. Potentially privileged documents move to a second review, if not more, for a final privilege determination. “It’s a slow, tedious, and expensive process,” Malhotra said.
Doug Austin, editor at eDiscovery Today, moderated the webinar, “AI & Privilege Review Today—and Tomorrow,” which featured a panel of litigation and eDiscovery professionals: Gareth Evans, partner at Redgrave LLP; Bobby Malhotra, eDiscovery counsel at Munger, Tolles & Olson; Joy Murao, founder & CEO of Practice Aligned Resources; and Jeff Grobart, associate director professional services at H5.
Holding on to current methods
Organizations continue to rely heavily on generic search terms and eyes-on reviews. The risks include over-capturing non-privileged documents and under-capturing privilege documents, which could lead to inadvertent production. Why do organizations continue to work this way? “As imperfect as the process is, it works,” explained Evans. “There’s a fear of trying something new when you have something that works fairly well, particularly given the stakes. For lawyers, the stakes in privilege review and privilege logging are very high. First of all, there’s an ethical duty that lawyers have to protect the privileged communications of their clients. Second, the most important and sensitive communications a business or individual has are communications with their attorney. There’s almost a stigma, probably rightfully so, if you inadvertently produce documents you should have protected.”
An issue that highlights the weaknesses of current privilege review workflows is the massive data volume that companies must address in discovery. Murao pointed out current methodologies work — until they don’t. With thousands or millions of documents in many reviews today, she warned about “data blindness.” Reviewers become blinded by repetitive information, the result being inconsistently coded documents. Again, the ultimate risk is inadvertent production of privileged documents, which forces firms to attempt to clawbacks.
Protection using Rule 502(d)
Here the panel took a moment to discuss the benefits of Federal Rule of Evidence 502, and in particular, 502(d), “which provides the court can enter an order that essentially puts into place a no-fault right to clawback,” explained Evans. “It does away with the disputes over whether you took reasonable steps or not. It simply provides, in its purest form, that if you inadvertently produce documents, you can claw them back.”
However, the safety of a 502(d) order, which attorneys are advised to use in every federal case, doesn’t mean organizations can avoid improving their privilege workflow methodologies. Grobart discussed the importance of improving keyword search functions beyond general terms. “Focus on privileged actors, privileged organizations,” said Grobart. “People who will break privilege are far more important to identify than what will capture privilege because being able to eliminate documents from your privilege review is just as important as capturing the potential privilege.”
Ways to improve
The panel next discussed several modern techniques that organizations can use at the beginning of the review to improve consistency, accuracy, speed, and costs. Name normalization allows for consistency with names throughout the review process and for the organization to become familiar with the players involved. Email stringing and de-duplication enable organizations to avoid reviewing the same or similar documents over and over. Email stringing also ensures reviewers have the context necessary to make an initial privilege determination.
Austin asked about the factors holding the legal industry back from adopting these and other technologies and approaches?
Malhotra said professionals, often lawyers, have to start believing and trusting that AI tools work. A lot has changed since 2015. The classification models available today are different and more accurate than TAR technology, which isn’t known for performing well in privilege reviews. Also, providers of these technologies have evolved. Not only do they offer better privilege identification models, but they also provide consultants who help organizations or firms tweak the classification model to their needs.
Grobart then discussed where AI and privilege reviews are headed. “Right now, what has really piqued my curiosity and makes me want to dive deeper is the thought of reinforcement learning,” said Grobart. “It’s the same technology people are using to create self-driving cars and teaching robots to walk. It’s really meant to mimic human behaviors and can be used to mimic human thought processes in a way that’s different than what folks might understand with existing deep learning models and other current technologies.”
The use of AI
The panel noted how they believe workflows are changing already due to new tools. Murao and Malhotra both spoke of fundamentally changing the first level of review with technology.
“I call that the holy grail of privilege review…” said Malhotra. “Taking the privilege determinations completely out of the hands of the L1 review team. There’s no decision-making if you’re running a TAR for responsiveness. There’s no privilege component to L1. Instead, using AI workflows and automated privilege determinations to replace that function. The thought being, and in my experience, that the AI tool will help you curate a more core set of likely privileged documents.”
Finally, the panelists made predictions on what the privilege review landscape might look like in the next few years. All believe more organizations and firms will adopt newer technologies for various reasons. As with any new technology, there will be some early adopters. As it becomes increasingly clear that AI technologies work, more professionals will feel comfortable using them.
“When the time comes, I think the AI tools that are being developed currently and in the future are going to be welcomed with open arms because it’s going to make everyone’s job easier,” said Grobart.
Watch the webinar, AI & Privilege Review Today—and Tomorrow, on demand here, part of the Working Smarter, Not Harder webinar series from H5. Additional resources are available on the webinar platform.