Anthropic, the US artificial intelligence start-up behind the Claude chatbot, has launched a new legal automation product that immediately unsettled markets, triggering a broad sell-off in listed information, analytics and software companies with exposure to legal and professional services.
Investors’ reaction was swift. Shares in Thomson Reuters fell sharply in US trading, while large European data and publishing groups also dropped, with moves in some cases reaching double digits during the session. The market response reflected concern that increasingly capable AI systems could compress margins for incumbent providers of legal research, compliance content, analytics and workflow tools that have historically relied on subscription revenues and high switching costs.
The scale of the sell-off was notable beyond the legal niche. Bloomberg reported that roughly $285bn of market capitalisation was erased across affected stocks, as investors reduced exposure to companies perceived to be vulnerable to “AI disintermediation” — the risk that customers bypass established intermediaries and buy capability directly from AI vendors.
What Anthropic launched is positioned as an expansion of Claude for workplace use. Reuters described upgrades that include plug-ins aimed at automating tasks across functions including legal, sales, marketing and data analysis. The legal component is intended to support routine corporate legal workflows, such as triage and review of contracts, compliance processes and internal briefings.
Anthropic has emphasised that the tool does not provide legal advice and is meant to be used under professional supervision. That caveat aligns with common constraints for legal technology products, which typically seek to avoid the regulatory and liability implications that follow from presenting outputs as advice to clients. Even so, the market’s response suggested investors were less focused on the fine print than on what the new product signals about the pace of automation and the potential for revenue substitution.
European-listed information businesses were prominent casualties. Reuters cited steep declines in RELX and Wolters Kluwer, both of which have built substantial legal and regulatory information franchises. The same report highlighted weakness across a wider set of names spanning data providers, legal services platforms and software groups. The Guardian also noted sharp moves in shares of several European publishing and data services firms, and reported that Thomson Reuters fell heavily.
In London, the declines fed into a broader market wobble. The Guardian reported that the FTSE 100 turned negative, despite having reached a record high earlier, as the AI-driven re-pricing spread through constituents linked to information and services.
The episode has renewed a familiar question: whether legal work, long viewed as resistant to automation, is now moving into the category of tasks that can be handled at scale by general-purpose AI systems. The current generation of models can draft and summarise, extract structured information from documents, compare clauses across versions, and produce templated responses. In-house legal departments, which already face pressure to cut external spend and shorten review cycles, have been active adopters of tools that promise time savings on repetitive work.
For incumbent vendors, the challenge is twofold. First, their value proposition often rests on proprietary content and trusted workflow integration. If a model can ingest a customer’s own document sets, generate workable summaries and propose edits without needing a traditional research database, the perceived necessity of certain subscriptions may weaken. Secondly, even where subscription products remain essential, pricing power may come under pressure if buyers believe the same outcomes can be reached with a lower-cost AI layer built on top of existing sources.
That does not mean the disruption is straightforward. Legal work involves judgement, risk tolerance and accountability, and large organisations will continue to require audit trails, controls and clear responsibility for decisions. AI outputs can be wrong, incomplete or difficult to defend, particularly where context is missing or where questions require interpretation rather than extraction. For many organisations, the near-term effect may be a reallocation of human time away from first-pass review towards supervision, escalation and final decision-making.
Even so, equity markets tend to price on trajectories rather than current practice. Reuters framed the sell-off as a reassessment of the “visibility premium” attached to software and data businesses with stable recurring revenues, as AI progress accelerates and lowers barriers for new entrants. The Times, in coverage of the UK market reaction, described sharp falls in valuation for companies with legal services exposure after the Anthropic announcement.
For policymakers and corporate leaders, the development adds to an ongoing debate about productivity, employment and professional standards in an AI-intensive economy. For the legal sector specifically, the central issue is likely to be governance: how firms and in-house teams document the use of AI tools, test outputs, manage confidentiality, and define responsibility where automated systems shape decisions. The market reaction this week suggests investors believe that, governed or not, the centre of gravity in legal technology may be moving faster than expected.
