The legal industry is reaching a point where adopting artificial intelligence is no longer the difficult part. Demonstrating that its outputs can withstand scrutiny is becoming a more pressing concern, particularly in litigation, where a flawed document review or an unreliable piece of evidence can have consequences well beyond efficiency.
That tension will be among the issues discussed at Georgetown Law’s Advanced eDiscovery Institute (AEDI), scheduled for November 19 at the Capital Hilton in Washington, D.C. The event expects more than 450 judges, litigators, in-house counsel and technology professionals, bringing together an industry that is increasingly relying on generative AI while still working out how to validate its use.
According to the Thomson Reuters Institute’s 2026 AI in Professional Services Report, 40% of professionals now work at organizations using generative AI, compared with 22% a year earlier. However, the infrastructure for evaluating its business impact has not kept pace. Only 18% of respondents said their organizations measure AI return on investment, while another 40% were unsure whether such assessments were taking place.
Client expectations are also evolving. Two-thirds of corporate respondents want their outside law firms to use AI, although fewer than one in five require them to do so.
The disconnect becomes more complicated when AI moves from administrative assistance into evidence handling and legal decision-making.
The challenge of making AI defensible in court
The debate over proposed Federal Rule of Evidence 707 illustrates how difficult it has been for the legal system to establish standards for AI-generated evidence. The proposal sought to subject AI-generated evidence presented without an expert to the same reliability requirements applied to expert testimony under Rule 702.
Following criticism from different sides of the legal profession, the Advisory Committee on Evidence Rules decided on May 7 not to advance the proposal or any of three revised versions. Instead, the committee allocated the first part of its October 15 meeting in Boston to a symposium involving technical experts and legal practitioners.
According to Redgrave LLP’s assessment of the process, August 2027 is now the earliest a potential amendment concerning AI-generated evidence could be published for public comment.
The relatively limited number of reported deepfake disputes has also influenced the discussion. A Federal Judicial Center survey received responses from more than 930 federal trial judges, but only 15 reported having faced a challenge alleging that evidence had been fabricated using AI.
For practitioners, this leaves a complicated environment: AI adoption continues to expand, but courts are still determining how existing evidentiary standards should apply to emerging technologies.
That uncertainty is particularly relevant to document review, where automation can accelerate the identification and analysis of information but does not eliminate the need for human oversight.
Shimmy Messing, CEO of Altorney, will be among the professionals attending this year’s institute. Founded in 2021, Altorney operates a marketplace connecting law firms, alternative legal service providers (ALSPs) and corporate legal departments with document review attorneys. The company also offers its MARC solution, positioning its operations close to the evolving eDiscovery landscape.
As legal teams introduce AI-assisted workflows, human reviewers remain part of the process, particularly when confirming privilege determinations, validating results and addressing potential errors. Their work also needs to be properly supervised and documented, especially when opposing counsel questions the methodology used during discovery.
This creates a broader operational challenge for legal departments and law firms: determining how human expertise should complement AI systems, how review processes should be structured and what those services should cost.
From AI discussions to practical testing
The 2026 AEDI program will also place emphasis on practical applications rather than limiting discussions to the opportunities and risks of emerging technologies.
Its AI Proving Grounds labs, hosted by Everlaw and Relativity, will return this year with a new feature: continuing legal education (CLE) credit. Participants will be able to work directly with the platforms on tasks such as early case assessment, fact development, chronology building and deposition preparation.
Both companies will also host pre-conference labs on November 18. These sessions, like the AI Proving Grounds activities, are available exclusively to in-person registrants.
The institute has been held for more than two decades, giving it a perspective on how legal technology has evolved. Technology-assisted review (TAR), once a contested approach to document analysis, eventually became an accepted practice in litigation.
Generative AI is following a different trajectory. Its adoption is advancing at a faster pace, while the legal precedents and standards governing its use remain comparatively limited.
For Messing and other attendees, the discussions in Washington will therefore extend beyond which platforms can automate the most work. The more consequential question is how legal professionals can build processes that combine technological capabilities with human judgment, measurable results and evidence that can withstand judicial scrutiny.
With more than 60 speakers on the agenda, AEDI arrives at a moment when the legal industry must move from experimenting with AI to establishing what responsible and defensible implementation looks like.
