LexisNexis Authorized & Skilled as we speak launched Protégé Normal AI, increasing its synthetic intelligence platform to incorporate safe entry to a number of general-purpose AI fashions alongside its present legal-specific AI instruments.
This eliminates the necessity to change contexts between legal-specific and general-purpose AI instruments and offers a better diploma of safety when utilizing the final instruments.
Launched as a preview model for U.S. clients, it permits authorized professionals to entry fashions together with OpenAI’s GPT-5, GPT-4o, and o3, in addition to Anthropic’s Claude Sonnet 4, all inside the firm’s present Lexis+ AI workflow. Customers can change between general-purpose and legal-specific AI with a toggle perform.
The brand new basic AI functionality works alongside Protégé Authorized AI, which makes use of a number of legal-tuned fashions together with GPT-4o, GPT-4.1, Claude 3.7, Claude 4, fine-tuned GPT-4o-mini, and fine-tuned Mistral. The authorized AI model grounds responses in LexisNexis authorized content material and validates citations by means of Shepard’s quotation service.
Protégé Normal AI attracts from open-web sources and is designed for duties together with analysis, drafting communications for each authorized and non-legal audiences, and breaking down complicated issues. The system features a Shepard’s Quotation Agent that checks authorized sources included generally AI responses.
Completely different Fashions, Completely different Strengths
Based on LexisNexis, every basic AI mannequin provides particular strengths when used for authorized work:
- GPT-5 combines basic and area information with reasoning and non-reasoning capabilities.
- Claude Sonnet 4 provides a pure fluency that gives well-structured responses for on a regular basis duties, communications and brainstorming.
- GPT-4o is sweet for on a regular basis duties similar to basic exploration, brainstorming, and web-based data integration.
- OpenAI o3 optimizes for deep analysis, strategic decision-making, and complicated problem-solving.
Agentic AI Framework
In the meantime, Protégé Authorized AI, constructed on the proprietary LexisNexis agentic AI framework, offers agentic workflows particularly developed for authorized work.
One among these, the AI Guided Analysis Workflow, consists of an Orchestrator Agent that breaks down complicated queries, a Authorized Analysis Agent that permits real-time consumer steerage, and a Reflection Agent that evaluations ultimate responses.
Some brokers perform as generalists for analysis and authorized questions, whereas others focus on particular duties like contract overview and quotation checking.
The system operates inside LexisNexis’s encrypted setting, which the corporate says offers enhanced privateness protections in comparison with client AI instruments. Organizations can disable basic AI entry solely for customers, and particular person customers can toggle the function on or off.
Sean Fitzpatrick, CEO of LexisNexis North America, UK, & Eire, stated the event responds to buyer requests for safe entry to general-purpose fashions whereas sustaining management over AI conduct inside authorized workflows.
Associated: On LawNext: How LexisNexis and Harvey Are Partnering to Reshape Legal AI, with LexisNexis CEO Sean Fitzpatrick.
“Our aim is to assist authorized professionals with a personalised AI assistant that permits uninterrupted workflows and entry to the complete vary of LexisNexis AI capabilities in a single safe place,” he stated.
Preview Program
Roughly 200 legislation companies, company authorized departments, and legislation colleges are taking part within the buyer preview program, which has already begun. Normal availability is predicted later this 12 months.
The event represents LexisNexis’s continued enlargement of AI capabilities, constructing on agentic AI innovation the corporate has been growing since 2024. The corporate says that it employs over 2,000 technologists, information scientists, and authorized specialists in its AI growth efforts.
LexisNexis companions with AWS, Anthropic, Microsoft, Mistral, and OpenAI for its multi-model method, which selects completely different AI fashions based mostly on particular use instances.
