プレッシャーの下で、LexisNexisがAnthropicの法律AIを統合

5 Min ReadJHVEPhoto via Getty ImagesAnthropic stunned the legal sector when it released a plugin for its Claude Cowork system in late January, triggering a dramatic stock selloff in LexisNexis, Thomson Reuters and other legal data vendors, and panic in an industry with particularly large exposure to AI.But rather than surrender or fight Anthropic head-on, LexisNexis adapted by incorporating the vendor's legal plugin technology into its Protégé platform.The legal information vendor said this is one example of how it's making the transition to an AI-dominated world by staying current with the latest AI advances and banking on what it says is a unique corpus of legal content on top of which the AI sits.In this Q&A with AI Business, Greg Dickason, executive vice president and CTO of LexisNexis, explains the company's AI strategy.Anthropic released its legal plugin for Claude Cowork on Jan. 30. You introduced the Lexis+ with Protégé platform with an Anthropic plugin integration on Feb. 24. How did you do that so fast?Related:The Real AI Shift Isn’t New Models. It’s Control.Greg Dickason: We were actually working with Cowork, Claude and Anthropic before they even went live with the plugin. That was quite a few years ago now, and as a result, we were close with them, and they tell us what's happening quite early. So, we got early access to Cowork, like a day's access, a day before they went live with the legal plugin. What do we do? We set up some tiger teams -- very, very rapid runners. They used agentic coding and were able to quickly build things, because we already had an agentic harness. With agents working inside our environment, it was quite easy to take new paradigms and ways of thinking and pop them in with a couple of clever developers very quickly.What was your thought process in arriving at the decision to incorporate rather than compete head-on with Anthropic?Dickason: If you think about where we're going, we're going to get lots of these sorts of skills and plugins coming from so many different places. Anthropic is the key one, but it's also going to come from ChatGPT and Google. If we can incorporate them, we can make a place for our lawyers to get maximum value. If there's some great workflow or some great skill that's been produced by Anthropic, why can't they use that in our environment, assuming we've got the legal signoffs, which in Anthropic's case, we do. We've also got a ton of content. We're a content business. We've been a content business for 100 years, and so we want to give our customers the best tech in the world. Trying to compete head-on, that doesn't make sense. This is a world of coopetition -- not competition, not direct competition.Related:OpenAI GPT-5.4-Cyber is More Open Than Claude MythosWhat does your contractual relationship with Anthropic look like?Dickason: We use Claude extensively in our system, even though we're multi-model. We have contracts with them because we use their API. But when a user uses our system, they don't know which model we're using because we're using multiple models at the same time.What's your view of the 'SaaSpocalypse,' the notion that generative AI will, if not kill traditional software, significantly take away its market share? What are some of the arguments against AI agents disrupting traditional software paradigms? And what are the arguments for it?Dickason: I think if you don't have a strong understanding of how users use you, and you haven't built up a good ecosystem, I don't think there's much you can do against AI. Because with vibe coding and agents, you can actually do relatively simple things very quickly. Where the defensibility comes in for your SaaS businesses is whether they've got really good content, like we do, or where they've got a deep ecosystem, where they're well integrated with lots of other players, and they're using and working with the customers' own content or data. And as a result, it's very hard to unplug that.Related:Anthropic Releases Good but not Great Claude Opus 4.7To what extent does AI have a credibility problem? If you are increasingly relying on it, which it seems you are, doesn't that make you liable for some of the credibility and safety risks that a pure legal AI tech platform could pose? And how does LexisNexis guarantee citation accuracy in general?Dickason: First off, we're always lawyerly. But we never get to the point where we are trying to give advice or anything like that. We've always been about giving the best possible content to the lawyer at the time when they need to make a decision, but ultimately, they're still the decision-maker. What we need to do is make sure we give them what we've grounded in all of our content, that we give them the easy ability to reference every citation we give them, and that the citations we give them are therefore applicable. You can go straight into the database and see them. We're never going to go direct to the customer. It's always about the lawyer in the loop.Companies such as OpenAI, Anthropic, Mistral and others are building increasingly powerful foundation models. What's your moat against them?Dickason: A couple of things. One is that they don't have content, and they haven't built the content the way we have. The other thing is, while we're using them, we're not changing whenever a new [model or agent] comes along. We can plug it in. We've got a quick way to say which is the best model, plug it in and as a result, we can use different models in the same workflow.  And I think our evaluation of models is pretty good.Let's hypothesize that the best legal reasoning is going to eventually come from AI models. You can disagree or agree with that, but if that is so, what role does a legal information provider play in that kind of stack?Dickason: Think of it like this. If I go and see a recent law school graduate or I go and see somebody who's got 20 years of experience in my particular area, I'm going to get very different reasoning from those two. One is going to give me generic thinking. The other deeply understands their domain. If I saw you over coffee and you've got 20 years of experience, you might give me a little bit of ad hoc advice. I can't see the model providers ever doing that. So that's why we add the content to the reasoning, and we get the best output.Editor's note: This interview was edited for conciseness and clarity.About the AuthorSenior News Director, AI BusinessShaun Sutner, a journalist with more than 25 years of daily newspaper experience and 11 years at Informa TechTarget as an editor and writer, directs news coverage for AI Business. He was previously a senior news and features writer covering health IT and HR software at TechTarget and a senior news director overseeing coverage of AI, business analytics, data management and government tech regulation.Sutner's newspaper career included investigative reporting and covering the Massachusetts State House and politics for the Worcester Telegram & Gazette. He has written about snow sports as a T&G columnist and correspondent for 20 years. Sutner's interests also include tennis, standup paddleboarding, cooking and popular music.