Every major AI lab has bet that frontier performance requires billions in compute. Thomson Reuters challenged that assumption.
The recording from our September 15 LinkedIn Live is now available on demand. CEO Steve Hasker and Alexander Kardos-Nyheim from TR Labs break down how Thomson — our first proprietary large language model — was built in-house for a fraction of typical frontier costs, and why owning the model matters when your customers depend on accuracy.
The session also covered why we chose to build our own model, what Fiduciary-Grade AI
actually means in practice, and what AI sovereignty looks like when the stakes are real.
If you missed it, the on-demand recording is now available. It's well worth the watch!