Product leads talk about why their model ranks first in class for legal factuality and gives firms a path to sovereign AI
The question of whether to buy or build AI is one that law firms across Canada are actively wrestling with. Thomson Reuters has now answered it for itself: the July 2026 launch of Thomson – its first proprietary large language model (LLM) – marks a strategic pivot away from relying solely on third-party models from providers such as OpenAI, Anthropic, and Google.
Two factors drove that decision. David Wong, the company’s chief product officer, and Alexander Kardos-Nyheim, who co-leads Thomson Reuters’ foundational machine learning research, discussed the shift in a recent CL Talk podcast interview. Kardos-Nyheim came to Thomson Reuters through the 2024 acquisition of Safe Sign Technologies, a legal AI startup he founded while completing his training contract at A&O Shearman.
Why Thomson Reuters built its own model
The Thomson LLM sits within what Wong calls a multi-model strategy – selecting the best available model for each task rather than committing to one provider. The first driver for building in-house is cost. “Using the latest frontier models is the shortcut to get good performance, but they’re very expensive,” Wong says. The second is data sovereignty. As concerns about where client data is processed grow, “the amount of trust with some of the big tech vendors is declining, and some, not all, ... firms are interested in rolling their own, running their own infrastructure, using their own tech, operating models by themselves,” he says. Thomson Reuters says it is exploring options to license the Thomson model to firms that want to run it on their own hardware without external providers, though it has not yet done so.
What domain-specific training delivers
To distinguish domain-specific training from retrieval-augmented generation (RAG) – the approach most legal AI tools rely on – Kardos-Nyheim uses a pointed analogy: “I’d compare domain-specific training to RAG as a closed book versus an open book exam.” RAG supplies the model with relevant documents at inference time. Training encodes knowledge directly into the model’s parameters – a fundamentally different level of capability.
The Thomson model was trained on Thomson Reuters’ content universe – Westlaw, Practical Law, and Checkpoint – and is trained on the work of thousands of lawyer-editors. On legal factuality benchmarks – the model’s ability to link a legal proposition accurately to the correct citation – the results stand out, Kardos-Nyheim says. “The model is not only first in class, but it is first in class by almost 20 points,” he says, attributing the margin to a focus on source hierarchy: “understanding that there is primary law that takes precedence.” He says the model has achieved this after seeing less than 10 percent of Thomson Reuters’ total content. “We have already managed to enhance the capabilities of this model significantly with just a small portion of what we think we have to offer,” he says.
Fiduciary-grade AI and the hallucination problem
For lawyers, “fiduciary-grade” has a specific meaning. Wong says it is a system built for professionals with fiduciary responsibility to their clients, where the consequences of error are greater and accuracy, privacy, and transparency are non-negotiable. “The burden of proof for truthfulness and accuracy is higher, and privacy, security, and transparency are higher,” he says. The framework rests on four principles: grounding in authoritative sources, built-in data privacy and security, human oversight incorporating professional expertise, and transparent and verifiable reasoning.
A recursive verification method – in which the system checks its own responses against cited sources to detect potential hallucinations – underpins the reliability architecture. “This recursive verification has largely eliminated the possibility of hallucination in our responses,” Wong says. Thomson Reuters’ lawyer-editors were also involved in a range of mid- and post-training tasks, including calibrating the model's judgment in complex legal skills such as legal research. Thomson's first production deployment is inside Tabular Analysis, a capability within CoCounsel Legal that allows bulk document review across thousands of files. CoCounsel Legal is Thomson Reuters’ agentic legal AI platform, with a workspace, skills, and workflows that give legal professionals native, real-time access to Westlaw and Practical Law. It's multi-model by design, drawing on several AI models depending on the task, with Thomson applied where its domain depth offers a clear advantage.
Buy or build – and what Canadian firms should ask
For law firms and in-house legal teams weighing AI strategy, Wong’s advice is to buy first. Most core problems – research, drafting, document analysis – are already well addressed by existing vendors. “For building, it really needs to be something which is unique … to the customer, to the problems, to the differentiation of the service,” he says. Where Thomson factors into the build decision is at the infrastructure layer, where firms have historically faced a forced choice between a sovereign model that may be less capable and a highly capable frontier model that routes data through third parties. “Our hope behind Thomson is that we no longer force customers to choose, that Thomson is a model that is both sovereign, that has that safety and security to it, and the frontier capability as well,” Kardos-Nyheim says.
Canada’s place in the rollout
Thomson Reuters is headquartered in Toronto, but the company has historically launched products in the United States first. Westlaw Advantage launched in the US in August 2025 and reached Canada in February 2026.
For skeptical lawyers who tried AI tools a couple of years ago and retreated after encountering hallucinated citations, Wong suggests they try again: “If you compare the performance of research systems today from even 6 months ago or 12 months ago, the quality has improved dramatically.” The company’s position is that AI research remains an interactive and hands-on process – a starting point to be verified with the full suite of Westlaw tools, not a replacement for them.
Kardos-Nyheim stresses the continued need for legal expertise: “Speaking as a lawyer myself, and indeed a skeptical one, I believe lawyers will still need to be actively involved in the practice of law… we will always still need to be involved and to own the response. The difference is how much work is involved in verifying what is put in front of you.”
Thomson Reuters is an Insights Partner at this year’s Canadian Legal Summit, which will bring together in-house counsel, private practice leaders, and legal tech innovators.
This article is based on an episode of CL Talk, which can also be found here:
The episode is also available on our CL Talk podcast homepage, which includes links to follow CL Talk on all major podcast platforms.
