Enterprises are collecting the right volume of data in only a reported 2 per cent of legal matters. AI prompts, system responses and AI-generated outputs are creating a new class of discoverable records that traditional discovery tools were not built to handle.
The finding appears in The State of Emerging Data 2026: AI Expands the Digital Universe. FTI Technology, the technology arm of consulting firm FTI Consulting, published the report in July 2026.
The report positions AI data as the fifth element in an emerging data paradigm. It joins shared access, chat messaging, linked content and dynamic document versions as a disruptor to established processes.
AI artefacts now appear in regulatory inquiries, litigation and internal investigations, often ahead of settled case law. Common challenges include distinguishing human-authored content from AI-assisted, AI-summarised or synthetically generated material, and updating retention settings and legal holds to preserve prompt histories, model identifiers and usage telemetry.
The report also warns that default system policies can purge AI artefacts before they are collected. Hallucinated assertions, once embedded in an email thread or summary pane, become discoverable records that may be mistaken for genuine viewpoints.
"AI will soon dominate emerging evidential trends," said Craig Hendley, a managing director at FTI Technology. "Enterprise copilots and standalone AI models are already generating relevant artefacts, prompts, autonomous actions and code output, that are landing inside investigative workflows."
Courts and rules are moving
Case law is beginning to catch up. In litigation in the US District Court in Connecticut, the court granted a motion to compel production of AI prompts used by the plaintiff's expert witness. It rejected the characterisation of those prompts as "notes". The court ordered the witness to disclose any AI prompts or queries and warned that sanctions were available.
The US Advisory Committee on the Rules of Evidence has also proposed a new rule, Rule 707 on machine-generated evidence. It addresses the reliability of technologies that generate predictions or inferences from data.
The Onna survey found the average organisation manages data from 10 collaboration platforms per matter, including tools such as Notion, Asana, Airtable and Monday.com. Collaboration data volume meaningfully drives litigation costs for two-thirds of organisations.
Source code is a growing exposure. A reported 37 per cent of organisations face intellectual property loss when an employee departs, and more than one-third of IP theft incidents now involve source code and algorithms. A further 37 per cent believe their collaboration data is more than doubling in volume year over year.
In FTI General Counsel Report, 63 per cent of legal department leaders said they were very or extremely concerned about the impact of emerging data sources on legal risk.
FTI recommends engaging emerging data and digital forensics experts early, mapping collaboration and project-management tools, and preparing proportionality arguments for discovery tied to AI system data. The full report is available at fticonsulting.com.