By Warren Parrino, Regional Vice President of Sales at Trustpoint.One
The past few years have seen an explosion in the usage of and discussion about Generative AI tools. The capabilities of Generative AI are indeed groundbreaking and offer tremendous potential across various sectors, but it is not suitable or necessary for every matter, and in many instances it might be compared to using a bazooka as a fly swatter. Below are several points to keep in mind when considering whether or how to use GenAi as part of your workflow:
-
Contextual Understanding Limitations:
Generative AI lacks, for the moment, the nuanced understanding of context that human professionals possess. In eDiscovery, legal issues often hinge on intricate details and subtleties of case law that AI may misinterpret or overlook. This can lead to incorrect conclusions or overlooked evidence. Courts have already seen examples of AI-generated legal content containing fabricated or inaccurate citations, most notably in Mata v. Avianca, Inc., where attorneys were sanctioned for submitting non-existent case law generated by AI. Additionally, research from organizations like the Stanford Institute for Human-Centered Artificial Intelligence has highlighted ongoing limitations in large language models’ reasoning and factual reliability.
-
Quality Control and Accountability:
Relying solely on AI for critical decision-making raises concerns about accountability. If an AI system makes a mistake in sorting Responsive from Non-Responsive documents, who is responsible? The lack of clear accountability could hinder the adoption of AI solutions in legal settings, where accuracy is paramount. The American Bar Association has emphasized that attorneys remain ultimately responsible for the work product generated with AI tools, reinforcing that human oversight is not optional but required under existing professional conduct rules.
-
Data Sensitivity and Security Concerns:
Legal teams handle highly sensitive information that requires robust confidentiality and security measures. Publicly available Generative AI systems often rely on large datasets, which can raise concerns about data privacy and potential breaches. This is particularly problematic in eDiscovery, where maintaining the integrity and security of evidence is vital. Regulatory frameworks such as the General Data Protection Regulation and guidance from bodies like the National Institute of Standards and Technology underscore the importance of strict data governance and risk management when deploying AI systems.
-
Integration Challenges:
eDiscovery processes are often deeply integrated with existing technological frameworks and workflows. The introduction of Generative AI may necessitate significant changes to these established processes. Organizations may face challenges (technological, financial, educational, etc.) in integrating AI solutions, leading to resistance from team members and clients accustomed to traditional and widely accepted methods. Industry analyses from firms such as Gartner consistently note that integration complexity and change management are among the top barriers to enterprise AI adoption.
-
Regulatory and Ethical Considerations:
The legal industry is heavily regulated, and the use of AI raises ethical questions regarding bias, decision-making transparency, and fairness in legal proceedings. Regulatory bodies are still grappling with the implications of AI, which may slow down its widespread adoption in legal contexts. For example, the proposed EU Artificial Intelligence Act and evolving guidance from the Federal Trade Commission highlight concerns around algorithmic bias, explainability, and accountability. The paucity of case law on many areas sure to be impacted by the use of GenAI should encourage legal teams to proceed with caution.
In light of just these 5 points (and there are many, many others that could be raised), it is essential that law firms, legal departments, and Alternative Legal Service Providers operate with guardrails in place to protect their clients and themselves from the risks associated with unsupervised GenAI tools.
At Trustpoint, we have continued to see extremely positive results from our hybrid approach, using tried and true methods of data filtering, culling, and volume reduction to shrink the document universe as much as possible first and then evaluating whether the matter is an appropriate use case for a GenAI tool. This approach aligns with best practices promoted by organizations like Electronic Discovery Reference Model, which emphasize defensibility, proportionality, and process transparency. We see the utility in these new tools and we value their efficiency, but we also understand that they still demand human oversight, involvement, and guidance.
In closing, while Generative AI has truly remarkable capabilities, it is essential to approach its implementation with a critical lens. The various limitations, accountability issues, and practical challenges suggest that it is not a one-size-fits-all solution, and its utility should be considered on a case-by-case basis in consultation with a team that has evaluated the tools, the circumstances, and knows the risks.