How law firms win the AI race: Alex Kardos-Nyheim on data, defensibility and staying power episode artwork

EPISODE · Feb 17, 2026 · 47 MIN

How law firms win the AI race: Alex Kardos-Nyheim on data, defensibility and staying power

from The Non-Billable Podcast · host Non-Billable

Alex Kardos-Nyheim is the founder of SafeSign Technologies and now co-lead of AI research at Thomson Reuters. Alex built a legal-specific large language model while still a trainee at A&O Shearman, ultimately selling the company in 2024. He explains why Safe Sign focused on building the “engine” rather than the interface and how a legal LLM was able to outperform some of the biggest AI labs on legal tasks.A central thread of the conversation is how the competitive landscape in legal AI has shifted. What began as a wave of “wrapper” products around foundation models is now evolving into a deeper contest over models, data and defensibility. Alex argues that the real moat lies not in user interface features, but in training models on high-quality legal data and building robust retrieval systems that reduce hallucinations and improve reasoning.We also explore how Thomson Reuters is thinking about the full stack: combining proprietary legal data, live market intelligence and a legal-trained model into a cohesive platform. Finally, Alex shares what he’s seeing inside law firms. Smaller firms should be using AI as a force multiplier to “punch above their weight”, while larger firms face a strategic imperative to convert their institutional know-how into machine-readable data to power their AI tools. His advice to managing partners is direct: hire data engineers, structure your data and lean into expertise. The Non-Billable Podcast is sponsored by Legora. To find out more about Legora, visit: https://legora.com/ Chapters00:01 Introduction01:20 Building a legal LLM while at A&O03:40 Engine vs wrapper: why Safe Sign focused on the model04:41 Life after the Thomson Reuters acquisition06:24 Inside Thomson Reuters’ “skunk works” AI team08:42 From point solutions to platform wars in legal AI11:49 The perfect stack: models, data and RAG15:07 Why retrieval engines matter more than you think18:33 Data as the moat23:56 Can small data beat big data?27:14 Making legal AI “market aware”31:05 The evolution of Co-Counsel33:57 What law firms are getting right and wrong about AI36:50 “Hire 30 data engineers tomorrow”38:22 Productising law firm know-how43:10 Tips if you started a legal AI company todayAbout Non-BillableNon-Billable is the media company for modern legal professionals across private practice, in-house and legal tech.Visit our website: ⁠⁠https://www.nonbillable.co.uk

NOW PLAYING

How law firms win the AI race: Alex Kardos-Nyheim on data, defensibility and staying power

0:00 47:08

No transcript for this episode yet

We transcribe on demand. Request one and we'll notify you when it's ready — usually under 10 minutes.

Frequently Asked Questions

How long is this episode of The Non-Billable Podcast?

This episode is 47 minutes long.

When was this The Non-Billable Podcast episode published?

This episode was published on February 17, 2026.

What is this episode about?

Alex Kardos-Nyheim is the founder of SafeSign Technologies and now co-lead of AI research at Thomson Reuters. Alex built a legal-specific large language model while still a trainee at A&O Shearman, ultimately selling the company in 2024. He explains...

Can I download this The Non-Billable Podcast episode?

Yes, you can download this episode by clicking the download button on the episode player, or subscribe to the podcast in your preferred podcast app for automatic downloads.
URL copied to clipboard!