Former Meta Researcher Launches Custom Legal AI Startup
The startup aims to help law firms build proprietary models to maintain control over internal data.
Updated on Oct. 22, 2026 in Artificial Intelligence
Live Poll
Do you trust specialized in-house AI models more than software licensed from large technology providers?
Former Meta researcher Ethem Coban has launched a legal tech startup focused on helping law firms build and customize their own internal AI models. Coban, who previously worked at Meta's Superintelligence Labs, is reportedly in talks to raise $6 million at a $25 million valuation.
Why it matters
Law firms are increasingly seeking to transform internal case files into proprietary competitive advantages. By shifting from off-the-shelf tools to custom-built models, these firms aim to secure full control over their internal data and minimize the risks of model hallucinations.
Large law firms are allocating significant capital to AI infrastructure, with Kirkland & Ellis committing $500 million to build proprietary software. This follows established investments by firms like Latham & Watkins, which is acquiring Nvidia servers to support in-house model customization.
The players
Ethem Coban
A former Meta Superintelligence Labs researcher specializing in legal AI and model hallucination reduction.
Meta
A major technology company building foundational large language models at its Superintelligence Labs.
Kirkland & Ellis
A global law firm that has committed $500 million to the development of proprietary legal software.
Latham & Watkins
A global law firm currently investing in Nvidia server infrastructure to support custom in-house AI models.
The details
Coban's startup focuses on adapting large language models to internal legal databases to improve answer accuracy. At Meta, he worked for two years training models to specifically mitigate hallucinations—a term for when AI systems produce false information with high confidence. His approach centers on training these models on private legal expertise rather than relying on generic public data models.
Timeline
Ethem Coban left Meta in October 2026.
The Tech Race
This venture signals a transition for the legal industry from merely testing consumer-grade AI toward building sovereign technical stacks. It competes against existing firm-led efforts to customize high-end hardware for internal data processing.
Law firms may soon deploy customized AI tools trained on their internal case files, likely changing how associates perform legal research and document review. The timeline for the release of these proprietary tools remains unannounced.
The takeaway
The move toward proprietary legal AI reflects a broader industry push for data sovereignty and accuracy over general-purpose model reliance. Watch for the official announcement of the firm's first major legal client or pilot program to determine if the customized approach improves research efficiency.
Further reading
Explore more developments in Artificial Intelligence regarding how private firms are adapting model training for internal data security.
Source note: This article includes information reported by Business Insider.
Live Poll
Do you trust specialized in-house AI models more than software licensed from large technology providers?
