Generative AI (GenAI) and large language models (LLMs) have proven to be a boon for the legal profession. Many law firms are now using this technology to work faster, scale their operations, and automate routine tasks. The result is law firms that are more agile and people-focused than ever before.
However, AI does come with significant challenges and limitations. To use the technology ethically and effectively, legal practitioners must be aware of these potential pitfalls and understand how to overcome them. In doing so, you can maximize the value you receive from AI. Here are some actionable tips.
1. Limitation: Hallucinations
Every AI model is trained on historic data. AI uses this data to learn how to recognize patterns, produce outcomes, and reason through problems. However, some datasets are outdated, flawed, or incomplete. These datasets and the nature of LLMs can lead GenAI to produce “hallucinations,” or false, misleading, or fabricated information. Hallucinations can pose serious problems for legal professionals, for whom accuracy and reliability are paramount to success.
Tip: Keep Humans in the Loop
Perform regular audits of your AI tools to identify potential hallucinations. If you are partnering with a vendor, ask whether their datasets are balanced and representative of the sample you are working with. Many AI tools now offer transparency and explainability, making it easier for legal teams to understand how the tool drew its conclusions.
At every stage of the process, keep a human in the loop, and always make sure human judgement is involved in critical decision-making. This is especially important in areas like criminal law, civil rights, and employment discrimination.
2. Limitation: Lack of Transparency
AI is often a black box. Legal professionals might struggle with what AI researchers call “explainability,” or understanding how AI arrived at a specific conclusion. In some cases, an LLM might produce responses that are based on case data but are not sourced to that data. This makes it hard to explain the results to a client or in court, potentially compromising accountability.
Tip: Invest in Explainability
Before implementing any AI tool, take the time to understand how it works. If you are working with a vendor, ask them to talk you through the process by which the algorithm produces results. When vetting AI tools, choose tools that offer visibility into the decision-making process. If you are using an LLM and you have control over the prompt, ask the LLM to explain its reasoning and provide sources, especially if it is relying on evidentiary data to draw conclusions.
3. Limitation: Regulatory and Ethical Concerns
Despite the widespread adoption of AI among legal professionals, regulatory and ethical concerns remain. Using AI for legal decision-making raises questions about data privacy, security, and confidentiality. Moreover, if an AI tool produces errors, who should be held responsible: the algorithm, the legal professional, or the vendor who developed the tool? These concerns can make some legal professionals reluctant to integrate GenAI and LLMs into their workflows.
Equally important, the legal profession requires empathy and compassion that a machine cannot replicate. In areas such as criminal defense, family law, and immigration, these values are especially critical.
Tip: Prioritize Accountability
It is the job of the legal professional to prioritize accountability. That means staying up to date with fast-moving ethical guidelines and compliance standards around the use of AI and doing your due diligence.
When onboarding a new vendor, ensure your new tools are compliant with applicable privacy laws in the US, like the California Privacy Rights Act (CPRA), and beyond, like Europe’s General Data Protection Regulation (GDPR), and take steps to ensure client data remains secure. Ensure that vendor agreements contain transparent guidelines around accountability and liability. And as always, keep a human in the loop: a member of your team should verify the accuracy of any AI-generated outputs before they are used in a case.
4. Limitation: Difficulties with Nuance
Interpretation and context are fundamental to the practice of law. However, GenAI and LLMs often find it difficult to parse the nuance inherent to a legal case. In cases where intent, interpretation, and precedent matter most, AI might deliver inaccurate results.
Tip: Think Augmentation, Not Replacement
AI is a tool, and like any tool, it is designed to augment human judgement—not replace it. AI excels at recognizing patterns and parsing unstructured data, which makes it a good fit for tasks like research or document review. For interpretive work, make sure a human is involved in the process.
While AI can make helpful suggestions, it cannot replace a legal professional’s empathy and expertise. After using AI to draft a memo, engage in document discovery, or carry out legal research, make sure there is a human responsible for reviewing, editing, and fact- and cite- checking that output.
5. Limitation: Dependence on Data Quality
AI is only as good as the data it relies on. This poses a potential issue for legal professionals. Legal data is often “noisy,” meaning it contains incomplete, extraneous, outdated, or inconsistent information. AI models trained on this data can produce inaccurate or biased results.
Tip: Take a Hands-On Approach to Data
Legal teams might be tempted to leave data to the data scientists. However, to reduce errors and mitigate bias, it is vital that you take a hands-on approach to data.
Fortunately, you do not have to have a background in data science to do so. Before integrating a new AI tool, work closely with the vendor to ensure their data is high quality. Ask them how frequently their databases are updated, whether the vendor is regularly checking for gaps in data quality, and how those gaps are addressed. Their data should be diverse, high-quality, and current. Always cross-reference AI-generated results with your own research.
Conclusion
AI can be a powerful tool for augmenting and streamlining your work. But to maximize the value you get out of AI, legal teams should adopt a skeptical, thoughtful approach. Use AI to accomplish tasks that require attention to detail and pattern recognition, but keep a human in the loop.
While efficiency is important, the legal profession also relies on very human skills such as empathy, intuition, and, most importantly, judgment. Centering those values is crucial for success as a legal professional.
