AI in E-Discovery: A Blessing Wrapped in a Dangerous Risk
In 2025, the legal world is witnessing a startling trend: lawyers submitting AI-generated fake case law to courts, a blunder that’s become alarmingly common. Many attorneys shrug these stories off, confident it won’t happen to them because they know better than to rely solely on AI for their legal research. But here’s where it gets controversial—the true threat might not be just the legal research itself, but the growing and often overlooked use of AI in the document discovery stage of litigation.
Lawyers Facing Consequences for AI Errors
This year, reports of lawyers—from small practices to large firms—misusing AI in their filings have flooded the news. Judges have responded by slapping sanctions on those submitting either fabricated or incorrectly cited case law. Penalties have varied widely, ranging from simple warnings to fines, and even to some creative solutions like ordering attorneys to send copies of the sanctions orders to all their clients, opposing lawyers, and judges involved in other cases. For instance, in Johnson v. Dunn (2025), a federal court took this unusual step. More strikingly, a California appeals court recently criticized opposing counsel for not flagging their adversaries’ AI-generated bogus case law, illustrating that the ethical duty to police AI misuse extends beyond just the offending party (Noland v. Land of the Free, L.P., 2025).
AI’s Potential in Revolutionizing Discovery
AI’s promise goes well beyond legal research into the discovery process, an area infamous for consuming too much time and driving legal costs sky-high. Anyone who has been through litigation knows that sifting through mountains of phone data, video calls, and even AI-generated prompts and outputs can be a nightmare. Here, AI tools offer a tantalizing solution: they can swiftly search for relevant information and even categorize documents automatically, determining what is responsive or protected by privilege.
However, the excitement surrounding these tools should be tempered with caution. Former Magistrate Judge Andrew Peck, an early advocate for technology-assisted review, famously cautioned that such tools are "not a magic, Staples–Easy–Button solution appropriate for all cases." Technology should be viewed as an aid rather than a replacement for human judgment, a sentiment still highly relevant as AI tools evolve.
Beware Blind Trust in AI Discovery Tools
Although AI has become embedded in many aspects of our everyday lives and society increasingly values its output, lawyers must resist the urge to place blind faith in AI when using it for document discovery. Ethical rules from the American Bar Association emphasize this clearly. For example, the comments to ABA Model Rule 1.1 require lawyers to stay informed about the benefits and risks tied to technology relevant to their practice. Moreover, ABA Model Rule 5.3 obliges attorneys to ensure that third-party vendors—like those providing AI discovery services—operate in alignment with the lawyer’s professional duties.
This means practitioners have a responsibility to understand how AI tools arrive at their conclusions, such as why a document is flagged as relevant or privileged. Most AI platforms provide explanations or highlight the text that influenced a decision, but lawyers must actively verify these outcomes to avoid mistakes.
The stakes are particularly high when AI is used to identify privileged documents. A failure by AI to detect such sensitive materials could lead to accidental disclosures, potentially breaching client confidentiality and causing serious legal repercussions. Hence, the best practice involves a second layer of human review of all documents flagged by AI, plus manual checks to ensure nothing relevant slips through the cracks.
Other Hidden Dangers
Incorrect document production isn't the only hazard. Lawyers must also scrutinize how client data is handled by AI vendors, especially concerning data privacy and security. For instance, if data is processed on servers based outside the U.S., compliance with privacy regulations such as HIPAA or GDPR can become complicated. Additionally, there’s a risk that client information might be used to train AI models, which could inadvertently expose confidential or privileged information in future AI outputs to unrelated parties.
Transparency Is Key When Using AI
For those integrating AI into discovery, transparency is crucial. In civil cases, early discussions about AI use should occur during initial meet and confer sessions with opposing counsel, and details about AI parameters should be codified within electronic stored information (ESI) protocols. In government investigations or subpoena responses, clear agreements regarding the AI process and oversight are equally important.
Laying out these AI-related procedures upfront and ensuring human oversight will help foster clear, defensible discovery processes and minimize risks.
Final Thoughts
In summary, the use of AI in e-discovery holds undeniable promise to reduce costs and speed up document review, but it’s a double-edged sword. Lawyers who fail to approach AI tools thoughtfully and responsibly risk serious ethical breaches, client harm, and negative headlines in the legal press. The question now is: are legal professionals ready to embrace AI with the necessary caution and transparency?
Is it possible to strike the perfect balance between efficiency and ethical responsibility in AI-assisted discovery, or will the legal field be forever haunted by AI missteps? Share your thoughts below!
Disclaimer: The views expressed here belong to the author and do not represent Reuters News or its commitment to impartial reporting. Westlaw Today is owned by Thomson Reuters but operates independently.
David A. Shargel is a partner at Bracewell LLP in New York, where he leads the ePractice department. He advises clients and colleagues on best practices around electronically stored information (ESI) and digital forensics, and represents companies and executives in government investigations and complex disputes involving agencies like the SEC and DOJ. Contact him at david.shargel@bracewell.com.
