Legal AI is splitting in two—and most people miss the difference | Fortune

Recent announcements from Thomson Reuters (CoCounsel reaching one million users) and Anthropic (expanded Claude plugins) highlight a crucial distinction in the legal AI landscape. The author argues that the market is splitting into two distinct categories: AI designed for operational legal tasks (like document standardization and workflow automation) and AI designed for authoritative legal work (requiring access to verified legal research, statutes, and regulations). A recent incident where Anthropic’s Claude pulled information from Wikipedia illustrates this point – it wasn’t an AI failure, but a ‘systems failure’ due to a lack of authoritative sources.

Anthropic’s new plugins target the operational side of legal work, directly competing with startups in that space. However, this does not equate to handling authoritative legal tasks, which require decades of curated legal data and expertise, like that provided by Thomson Reuters’ Westlaw and Practical Law. The author emphasizes that simply improving foundation models isn’t enough; a robust infrastructure and governance system are essential for reliable, high-stakes legal work.

The key takeaway is that the value and risk associated with different types of legal work necessitate different AI solutions. While both operational and authoritative AI have a place in the market, confusing the two obscures the true competitive dynamics and the importance of authoritative content and expertise in delivering trustworthy legal AI solutions.

(Source：Fortune)