This section introduces three categories of generative AI tools that law students and researchers may encounter in academic or professional settings. Whether you're just beginning to explore these tools or already using them in your studies or work, this guide provides context for understanding how they function and what they can and cannot do.
As of late 2025, the legal technology market has exploded with more than 500 generative AI tools, including more than 100 promoting "agentic" features. This guide is not intended to cover them all. Instead, it focuses on tools that (1) UC Davis Law students can access, (2) offer free or discounted educational access, or (3) are generating significant discussion in the legal field.
The categories used in this guide -- foundational, specialized, and agentic -- are a framework. Many tools blur these lines, and vendors often rebrand as they add new features. Agentic features, such as planning, adapting, or performing multi-step tasks, are being widely promoted as the newest development. While these claims often run ahead of reality, the concept is shaping how companies describe their products.
Foundation models are large, general purpose AI systems trained on vast datasets that include internet content and publicly available sources such as books and articles. Tools like ChatGPT, Claude, and Gemini can generate text, summarize content, translate language, and answer questions across many topics. Some models can generate citations or references when asked, but these may not always be accurate, complete, or drawn from reliable sources. Both the outputs and any cited materials should be verified against trusted sources rather than treated as a definitive answers.
| Tool | ChatGPT (OpenAI) |
| Description |
ChatGPT, first launched in 2022 by OpenAI, uses large-language models in the GPT (Generative Pre-trained Transformer) series to predict text based on patterns learned from massive datasets. It can draft documents, summarize materials, answer questions, translate text, and assist with research and writing. In 2025, OpenAI introduced the GPT-5 generation as the flagship model, with further refinement in the November 2025 announcement of GPT-5.1, which features two variants: GPT-5.1 Instant (optimized for conversational speed and clarity) and GPT-5.1 Thinking (optimized for deeper reasoning and more complex tasks). OpenAI’s o-series models (including o1, o3, and o4-mini) devote more computational steps to complex tasks, allowing them to follow longer instructions, evaluate more variables, and work through multi-part prompts with greater stability. In ChatGPT, they can also use built-in tools such as web search or Python when appropriate. OpenAI has also introduced “agent” functionality allows ChatGPT to carry out multi-step tasks using tools such as web browsing, file analysis, code execution, and workspace features for creating documents or other outputs. |
| Access | Free tier: provides basic access to ChatGPT’s current flagship model with usage limits such as message caps, slower response times, and potential wait periods during high-traffic hours. ChatGPT Plus ($20/month): offers faster responses, higher usage limits, priority access during peak times, and access to additional features such as file uploads and select tool or agent functionality. ChatGPT Pro ($200/month): designed for heavy or professional use; provides access to the most capable models available in ChatGPT, expanded usage capacity, and support for more advanced features, including complex agent-based workflows. Note: Model availability and feature sets may change; users should consult OpenAI’s current pricing page for the most up-to-date details. |
| In Practice |
A March 2025 article in Law Technology Today surveyed lawyers using ChatGPT for tasks such as document drafting, summarizing, client communications, legal research, etc., while also emphasizing the need for oversight and guarding against risks like bias, hallucinations and confidentiality breaches. In February 2025, lawyer and journalist Bob Ambrogi tested ChatGPT's Deep Research feature by asking it to assess the legality of the Trump administration's pause of federal grant, loan, and other financial assistance programs. Within ten minutes, the tool generated a 9,000-word memo outlining legal authority, potential legal challenges, case law, grant recipients' rights and remedies, and arguments for and against the policy's legality. It relied primarily on publicly-available sources. Ambrogi questioned how the tool might perform with access to proprietary legal research platforms like Westlaw or Lexis+. Read the full analysis on LLRX. In a separate comparison of OpenAI's Deep Research, Lexis+ AI, Westlaw Precision AI, and Vincent AI in February 2025, Ambrogi evaluated responses to a query about class certification appeal standards under California law. The focus was on the "death knell doctrine," which requires immediate appeal of class certification denials. Deep Research correctly explained the doctrine, Lexis+ AI omitted it, and Westlaw Precision AI returned a potentially confusing warning. Among the legal-specific tools Ambrogi tested, he found Vincent AI's answer the most complete. In May 2025, practicing attorney Carolyn Elefant described using ChatGPT’s Deep Research mode to draft a research memo on whether temporary utility entry onto property constituted a taking. ChatGPT summarized case law, Maryland statutes, and comparative authority. Andrew Perlman's 2024 article in Michigan Technology Law Review pointed out that ChatGPT might increase legal services access (e.g., lower cost, faster intake) but that it also raises questions about accuracy and professional responsibility. |
| More Info |
OpenAI’s ChatGPT Release Notes provide current feature updates, and its ChatGPT pricing and model comparison page outlines differences between free, Plus, and Pro plans. |
| Tool | Claude (Anthropic) |
| Description |
Claude is an AI assistant developed by Anthropic for text generation, document analysis, and multi-step task support. It can summarize, draft, classify, and interpret documents across long contexts. In 2025, Anthropic introduced a preview feature enabling Claude to create and edit files such as Excel spreadsheets, Word documents, PowerPoint slide decks, and PDFs directly within Claude.ai and the desktop app for higher-tier subscribers. Additional features include a browser-based coding environment (“Claude Code on the web”) and project-scoped memory options, with newer Claude 3.5 and 4.x model updates aimed at improving stability and performance in extended tasks. |
| Access | Free tier: provides basic access with usage constraints such as lower message quotas, slower response times, and potential wait periods during high-traffic hours. Claude Pro ($20/month): individual subscription with increased usage capacity, priority access during peak times, and early access to new features. Claude Max: higher-capacity tiers offering substantially expanded usage limits (e.g., approximately 5× or 20× Pro), priority access, and previews of advanced features; pricing typically around $100/month or $200/month depending on usage needs. Note: Model availability and feature sets may change; users should consult Anthropic’s pricing page for current information. |
| In Practice |
An October 2025 blog article from Clio explains how Claude is being adopted in law firms for drafting, research, contract review, and document workflows, but also emphasizes the need for oversight, verification, and attention to professional responsibility and confidentiality. In an April 2025 article, researchers described evaluating Claude using a benchmark based on questions from the Brazilian bar exam. Claude completed the questions under exam-like conditions and achieved the highest average score among the models tested. In a UK legal-analytics study in May 2024, Claude was used to classify summary-judgment decisions by topic, a task requiring interpretation of judicial reasoning due to the absence of built-in topical metadata. Claude reached an accuracy of roughly 87%. In an informal evaluation in March 2024, a law librarian prompted Claude to draft a legal memo from a hypothetical fact pattern. The model produced a structured draft with citations, and the evaluator noted that human verification of sources and analysis remained essential. |
| More Info |
Anthropic’s official Claude webpage provides current model information and feature updates, and its pricing page outlines differences among Free, Pro, and Max subscriptions. |
| Tool | Microsoft Copilot (Microsoft 365) |
| Description |
Microsoft 365 Copilot is a generative-AI assistant embedded into Microsoft 365 applications such as Word, Excel, PowerPoint, Outlook, and Teams. It supports productivity tasks including drafting emails, summarizing meetings, creating presentations, formatting documents, and analysing data. Copilot draws on large-language models developed by Microsoft and OpenAI and adapts to user context within the Microsoft 365 environment. Latest feature updates include enhanced agent capabilities, extensibility in Copilot Studio, and new survey/“@mention” features in Excel and mobile apps. |
| Access |
For UC Davis students, faculty, and staff, Microsoft 365 Copilot Chat is available free of charge at https://m365.cloud.microsoft/chat. Login requires UC Davis credentials and authenticates through Microsoft Entra ID. Using Copilot without this authentication means that data entered into Copilot may not be securely stored or protected. See the Copilot Product Brief from UC Davis Information and Educational Technology (IET) for more information. Paid tiers: Enterprise Copilot services that integrate into Microsoft 365 apps are available at additional cost through Microsoft but are not included in the campus license. |
| In Practice |
In April 2025, Microsoft highlighted Copilot use by Mike Morse Law Firm in Outlook and Word, combined with secure data sources, to draft legal documents. A March 2025 qualitative interview study showed that Copilot helped with routine work like meeting notes, email drafting, and summarization, but was less effective for complex reasoning or highly specialized workflows. In November 2024, global law firm DLA Piper reported using Microsoft 365 Copilot to streamline document creation, optimise workflows, and save up to 36 hours weekly on content generation and data analysis. In June 2024, Microsoft reported that global law firm Clifford Chance has adopted Copilot to streamline compliance and workflow documentation. |
| More Info |
The Microsoft 365 Copilot Release Notes page gives up-to-date feature and version information. Microsoft’s “AI for Legal Research: Tools and Best Practices” and "Using Microsoft 365 Copilot in Legal Departments" pages provides an overview of legal department use cases and governance considerations. |
| Tool | DeepSeek (Hangzhou DeepSeek Artificial Intelligence Basic Technology Research Co., Ltd) |
| Description |
DeepSeek is a Chinese AI company founded in 2023 that develops open-weight large language models for reasoning, text generation, and multilingual tasks. Its DeepSeek-R1 model attracted significant public attention when it was released in early 2025. Research evaluating R1 notes that while it can perform competitively on certain reasoning benchmarks, results vary depending on evaluation design, and independent reproducibility studies emphasize caution when interpreting comparative claims. |
| Access |
DeepSeek’s chatbot is available through chat.deepseek.com and via mobile applications. |
| In Practice |
A May 2025 report described DeepSeek's incorporation into legal practice and judicial workflows in China. Large firms, including Yingke Law Firm, have integrated DeepSeek-R1 into internal systems for regulatory research, case library search, contract services, and client matter intake, supported by domain-specific tuning of the model. Judicial institutions have piloted DeepSeek use, and in March 2025 China’s Judicial Convenience Platform added DeepSeek to provide online legal-consultation services. A March 2025 post on the Criminal Law Library Blog, Evaluating DeepSeek for Legal Research: Capabilities, Risks, and Comparisons explored DeepSeek’s application in legal research. The article noted that while the tool is fast and multilingual, its limited citation transparency, jurisdictional accuracy, and regulatory risks may pose challenges for U.S. legal professionals. In a February 2025 technical blog post, SkyPilot detailed a pilot retrieval augmented generation (RAG) system built using DeepSeek R1 for legal document workflows. They found R1 performed poorly at retrieval tasks but strong at generation. In China, DeepSeek has been used in public administration. As of early 2025, DeepSeek‑R1 was locally deployed for document drafting and policy interpretation in several municipal government units (e.g., Shenzhen, Foshan, Beijing), with “AI civil servants” assisting across districts. Also in early 2025, governments in the U.S., Australia, Taiwan, South Korea, and Italy imposed restrictions or bans on the use of DeepSeek due to concerns about national security and data privacy. Some U.S. federal agencies have blocked use on government devices, and legislators have introduced bills in the House and Senate to formalize broader restrictions. In April 2025, a U.S. House select committee report warned of potential ties between DeepSeek’s backend infrastructure and Chinese military-linked firms. |
| More Info | For updates and model access, visit the DeepSeek website, Hugging Face profile, or GitHub repository. |
| Tool | Gemini (Google DeepMind) |
| Description |
Gemini is Google DeepMind’s family of multimodal AI models capable of processing text, images, audio, video, and code. Gemini integrates across Google products, including Gmail, Docs, Sheets, Drive, Search, Maps, and Android, and is accessible through the Gemini web interface, mobile apps, Google AI Studio, and Vertex AI . |
| Access |
For UC Davis students, faculty, and staff, Gemini for Education is included free with Google Workspace accounts. The educational version includes enterprise-grade data-protection assurances (chat/files not used to train Google’s generative models) and defined usage limits (e.g., 1,000 uses/month, 10 Deep Research reports/month). Free tier: users can access Gemini at gemini.google.com with a Google account; availability of features varies by region and usage limits apply. |
| More Info |
See the Gemini Product Brief from UC Davis Information and Educational Technology (IET) for local guidance. For official model updates, release information, APIs, and subscription options, visit Google DeepMind’s Gemini page or Google AI Studio. |
| Tool | Grok (xAI) |
| Description |
Grok is a conversational LLM-based assistant developed by xAI (founded by Elon Musk). It can access real-time information from the web and the social-media platform X (formerly Twitter). It is designed for a range of tasks: answering questions, brainstorming, coding, generating documents, and, via add-ons, image/video generation. On November 17, 2025, Grok announced that its latest LLM, Grok 4.1, is available to all users. The name “Grok” draws from a verb coined by Robert A. Heinlein in his science-fiction novel Stranger in a Strange Land, meaning "to understand profoundly and intuitively." |
| Access |
Free tier: Provides limited access to chat models. Available via the Grok website/app and via X for certain user-tiers. Note: Model availability and feature sets may change; users should consult Grok’s current pricing page for the most up-to-date details. |
| In Practice | National Law Review’s AI Governance Series uses Grok as a case study in how system prompt changes can lead to problematic or extremist outputs. |
| More Info | xAI's "Latest News" page provides Grok updates, and its pricing page outlines differences between free and paid tiers. |
| Tool | Llama (Meta AI) |
| Description |
Llama is Meta’s family of large-language models designed for general-purpose text generation, reasoning, coding, and, in recent versions, multimodal tasks involving text, images, audio, and video. In April 2025, Meta released Llama 4, introducing variants such as "Scout" and "Maverick," with a larger model called "Behemoth" in development. Llama models are distributed under Meta’s source-available license and frequently appear behind third-party tools, hosted APIs, and research platforms. |
| Access |
Llama models can be used through Meta’s AI assistant at meta.ai. |
| In Practice |
A September 2025 multilingual legal-reasoning study found that Llamas performance on legal tasks (including classification, summarization, open questions, and general reasoning) lagged behind its performance on general purposes tasks. A 2024 study reported that Llama-2 hallucinated 88% of the time when answering a direct, verifiable question about a randomly-selected federal court case. A 2024 experiment using a Llama-3 model fine-tuned to enhance its legal text processing capabilities demonstrated improved performance over baseline models. |
| More Info |
For the latest version information, model cards, and licensing details, see Meta’s official Llama page. Model weights and implementation tools are also available on GitHub and Hugging Face. |
| Tool | Mistral AI |
| Description |
Mistral AI is a French company founded in 2023 that develops open-access large language models designed for efficiency, speed, and competitive performance. Its open-weight and open-source models are used in research, commercial applications, and multi-model platforms. In June 2025, Mistral introduced Magistral, its first reasoning model. |
| Access |
Users can try le Chat, Mistral's AI assistant, at chat.mistral.ai. |
| In Practice |
Harvey, a generative AI platform designed for the legal industry, announced a strategic partnership with Mistral AI in May 2024. |
| More Info | For model cards, release notes, and licensing details, visit https://mistral.ai or https://docs.mistral.ai. Model weights and example implementations are available on Hugging Face. |
| Tool | Qwen (Alibaba) |
| Description |
Qwen is a family of LLMs and multimodal models developed by Alibaba Cloud, originally introduced in 2023 under the name Tongyi Qianwen. The models support writing, research assistance, coding, multilingual tasks, long-context processing, and multimodal interpretation. The current flagship generation, Qwen3, was released in April 2025. Qwen3 introduced a broad model lineup, including a "hybrid reasoning" design that allows the model to shift between a slower, deeper reasoning mode and a faster conversational mode depending on task demand. Later in 2025, Alibaba expanded the lineup with Qwen3-Omni, a multimodal model capable of handling text, images, audio, and video, and Qwen3-Max, a model optimized for complex and large-scale workloads. |
| Access |
Users can interact with the models directly through Alibaba’s chat interface at chat.qwen.ai. Qwen's Research page provides model updates. Most Qwen models can be freely downloaded and self-hosted through platforms like Hugging Face, GitHub, and Alibaba’s ModelScope. |
| More Info | You can find more information about Qwen models, including technical details and download links, on GitHub, Hugging Face, or the official Qwen website. |
Some AI models are marketed as having reasoning capabilities (performing multi-step analysis via chain-of-thought) or with deep research modes (blending generation with real-time document retrieval and citations).
However, these advanced features do not guarantee better accuracy. In fact, hallucination rates increased in reasoning models, according to OpenAI’s own tests, as reported by The New York Times in May 2025:
Two more recent studies exposed deeper vulnerabilities:
These findings highlight a familiar limitation or generative AI: even models with chain-of-thought or deep research capabilities can be unreliable. In legal contexts, accuracy and valid sourcing are critical, and errors in these areas often have significant consequences. Users must independently verify outputs, particularly reasoning steps, and be wary that tools may appear confident while compounding errors under the surface.
Agentic models are an emerging class of AI systems designed to perform multi-step tasks with limited autonomy. Unlike generative tools that respond to single prompts, agentic models can plan sequences of actions, adjust based on intermediate results, and interact with tools or documents to pursue a defined goal. The shift is from one-off assistance (e.g., “summarize this case”) to broader task support (e.g., “draft a summary, extract key holdings, and prepare a comparison chart”).
Common features include:
That said, the term “agent” is being used in wildly inconsistent ways across the tech industry. As this TechCrunch article bluntly points out, there’s no shared definition of what counts as an AI agent. Some tools labeled as agents run a series of scripted steps. Others make dynamic decisions, use multiple tools, or act on the user's behalf.
At Legaltech Hub’s April 2025 Legal Tech & Innovation Conference, Dan Hoadley, Head of R&D and vLex Labs at vLex, offered a working definition of agentic AI. A true agent, he proposed, must reason about tool use, create and adjust a plan in real time, and know when its goal has been achieved. By that definition, most tools marketed as “agentic” in legal tech today fall short.
Legal tech companies like Harvey, Paxton, and LawDroid have started integrating agentic components into tools for regulatory mapping, contract analysis, and cross-jurisdictional research. These tools still require user input and oversight, but they mark a shift in how AI can support legal workflows.
As with many AI developments, the hype around agentic models has grown fast. Some marketing claims outpace the tools' actual capabilities. It’s worth asking what “agentic” really means in context and whether a tool’s features support that label. Still, the technical progress is real, especially for students and researchers exploring automation, legal operations, or AI-assisted legal practice.
These AI tools have been customized for legal tasks, typically using two main approaches: fine-tuning and retrieval-augmented generation (RAG). Proprietary generative AI tools like Protégé/Lexis+ AI and CoCounsel use one or both of these strategies. They tend to perform better in legal contexts than general-purpose AI tools and can help with research, drafting, and summarizing. However, their output still requires human review.
Fine-tuning means the model has been retrained on legal information (e.g., case law, statutes, or legal writing) so it “learns” patterns and vocabulary from within the legal field. This process permanently alters the model to make it better suited for legal applications.
RAG, by contrast, does not retrain the model. Instead, it allows the AI to pull relevant information at runtime from a trusted external source (such as a legal research database) before generating a response. This is intended to improve accuracy and reduce hallucinations, especially when the retrieved content comes from authoritative sources of law.
| Tool | CoCounsel |
| Description |
CoCounsel is a generative AI tool developed by legal tech company Casetext, launched on March 1, 2023, and initially powered by OpenAI's GPT-4. After Thomson Reuters (the parent company of Westlaw) acquired Casetext in August 2023, it expanded CoCounsel's capabilities. Thomson Reuters integrated CoCounsel into Westlaw Precision, its flagship legal research platform, in November 2023. When CoCounsel debuted in Westlaw Precision, it included eight "skills," including AI-assisted research and document review, summarization, and drafting. In June 2025, Thomson Reuters announced it would integrate agentic AI into CoCounsel. In August 2025, it launched CoCounsel Legal, which can execute multi-step tasks autonomously and includes built-in checkpoints for user oversight. |
| Access |
UC Davis Law students can access CoCounsel by logging into Westlaw Precision and selecting "CoCounsel" from the top menu or by selecting CoCounsel links on resource-specific pages. Law students and faculty gained access to Search & Summarize Practical Law (formerly Ask Practical Law AI) in January 2024. Law students and faculty gained access to other CoCounsel features in January 2025. |
| In Practice |
CoCounsel supports a range of legal tasks:
|
| More Info | For additional resources, including tutorials and product updates, visit the GenAI Resources page on Thomson Reuters' website. |
| Tool | Protégé (Lexis+ AI) |
| Description | Lexis+ AI is an AI-powered legal research and drafting platform developed by LexisNexis. It integrates generative and agentic AI technologies to assist users in conducting research, drafting documents, and analyzing legal information. The platform features Protégé, a personalized AI assistant. |
| Access |
UC Davis Law students can access generative AI tools in Lexis by logging into Lexis+ and selecting "Protégé" from the left-hand sidebar. Lexis+ AI launched commercially in the U.S. on October 25, 2023, and became available to ABA-accredited law schools in December 2023. The Protégé assistant was introduced in commercial preview in August 2024 and became available to law students in February 2025. |
| In Practice | Protégé offers features such as conversational search, legal document drafting, summarization of legal texts, and document upload and analysis. Users can upload and save documents to enable tasks like summarization, drafting, and research on the uploaded content. While the platform is grounded in Lexis’s proprietary legal content, it is not error-free. A 2024 study by researchers from Stanford's Regulation, Evaluation, and Governance Lab (RegLab) and Institute for Human-Centered Artificial Intelligence (HAI) found that an earlier version of Lexis+ AI’s answers were accurate for 65% of queries. Outputs should always be reviewed and verified using primary sources. |
| More Info |
Visit the Lexis+ AI Practice Page for FAQs and video tutorials. A law librarian’s critique offers insight into the platform’s “ask a legal question” feature, including examples of where it performed well and where it struggled. |
| Tool | Bloomberg Law Answers |
| Description | Bloomberg Law Answers integrates with Bloomberg Law Search to provide AI-generated responses to legal questions. The answers, which appear at the top of search results, include citations and links to Bloomberg Law's primary and secondary resources. |
| Access |
Bloomberg Law Answers launched on January 14, 2025. The tool is embedded directly within the Bloomberg Law platform and appears at the top of search results. No separate login or interface is required. Current UC Davis Law students can access Bloomberg Law through the law library's A-Z Databases page. Students who have not yet registered for Bloomberg Law should contact the reference desk for an individual registration code. |
| In Practice | The Bloomberg Law Answers tool is integrated into the Bloomberg Law platform and is currently in beta. Ongoing refinement is expected. |
| More Info | For further details about Bloomberg Law Answers and its integration into the legal research workflow, refer to the article "Bloomberg Law Introduces Next-Gen AI Tools for Legal Professionals." |
| Tool | Bloomberg Law AI Assistant |
| Description | Bloomberg Law AI Assistant is chat-based tool designed to generate summaries of documents and answer user queries about specific aspects of a document. |
| Access |
Bloomberg Law AI Assistant launched on January 14, 2025. The AI Assistant is integrated into the Bloomberg Law document viewer. When reviewing a case, regulation, or other legal document, students can activate the assistant directly from the document screen. Current UC Davis Law students can access Bloomberg Law through the law library's A-Z Databases page. Students who have not yet registered for Bloomberg Law should contact the reference desk for an individual registration code. |
| In Practice | The Bloomberg Law AI Assistant is integrated into the Bloomberg Law platform and is currently in beta. Ongoing refinement is expected. |
| More Info | For further details about Bloomberg Law AI Assistant and its integration into the legal research workflow, refer to the article "Bloomberg Law Introduces Next-Gen AI Tools for Legal Professionals." |
| Tool | Descrybe.ai |
| Description | Descrybe.ai is a free, AI-powered legal research platform that uses generative AI to summarize and simplify judicial opinions from U.S. state and federal courts. Designed to improve access to legal information, it provides plain-language summaries of over 3.6 million decisions and supports both legal professionals and the public. |
| Access | Descrybe.ai is publicly available online at descrybe.ai with no registration or subscription required. Users can enter natural-language questions and browse simplified summaries in English or Spanish. |
| In Practice | Students and legal professionals can use Descrybe.ai to quickly review case summaries, find relevant decisions, or understand precedent without reading full opinions. The platform includes tools to search by issue, keyword, or fact pattern. It also offers simplified summaries to support users with limited legal or reading literacy. |
| More Info | For additional details on Descrybe.ai’s goals and development, visit the About page. A full FAQ is available here, including information on coverage, accuracy, and ethical use. |
| Tool | Harvey (Counsel AI Corporation) |
| Description | Harvey is a generative AI platform built on custom-trained language models for legal work. It supports contract analysis, due diligence, litigation support, and regulatory compliance. Harvey is deployed as an enterprise tool through Microsoft Azure and APIs, emphasizing security, confidentiality, and "agentic workflows" for law practice. In 2025, launched Ask LexisNexis, integrating primary law, Shepard’s, and Lexis+ AI tools into its platform. |
| Access |
Enterprise / Commercial: Initially available only to major firms, Harvey is now broadly accessible via Microsoft Azure Marketplace for organizational customers. Educational: In August 2025, Harvey launched a Law School Alliance Program, providing free access to its platform for participating schools. The inaugural partners include Stanford, NYU, Michigan, UCLA, University of Texas, and Notre Dame. The program supports faculty in building AI-related curricula and lesson content during the 2025–26 academic year. Students and administrators at participating schools can also use Harvey at no cost. Harvey staff emphasized that the partnerships will be tailored to each school’s curriculum and faculty priorities, with the aim of developing “AI fluency” and practical prompting skills. |
| In Practice |
Allen & Overy (now A&O Shearman) was an early adopter, with ~4,000 users across 43 jurisdictions using Harvey. Other adopters include Ropes & Gray, Paul Weiss, PwC Singapore, and Ashurst. |
| More Info | For more information, visit the Harvey website. A detailed overview of Harvey's capabilities is available on Clio's blog. |
| Tool | midpage.ai |
| Description |
midpage.ai is an AI-enabled legal research platform covering U.S. federal and state case law, including administrative decisions from the Merit Systems Protection Board and Equal Employment Opportunity Commission. It displays results in a grid format for side-by-side comparison of legal issues and treatments. Users can generate AI-based case summaries, submit custom prompts across documents, and conduct proposition-based searches to find cases supporting specific legal principles. It includes an AI-powered citator and integrates with Microsoft Word. |
| Access | UC Davis Law students can access midpage.ai through individual subscriptions. The platform offers a Law Student Package at $10 per month. A two-week free trial is available for new users. |
| More Info | For additional information, including tutorials and support, visit the midpage.ai website. A detailed review of midpage.ai's features and performance is available on Artificial Lawyer, providing insights into its capabilities and applications. |
| Tool | Spellbook |
| Description |
Spellbook is a generative AI tool that uses large language models to assist lawyers with drafting and reviewing legal documents. Integrated directly into Microsoft Word, it offers features such as instant redlines, clause drafting, detection of unusual terms, and identification of missing or nonstandard boilerplate. On August 22, 2024, Spellbook announced the release of Spellbook Associate, an AI agent designed to handle complex, multi-step workflows in transactional matters. According to the company, Spellbook Associate performs tasks similar to those of a junior associate, such as drafting financing documents from term sheets, reviewing documents for risks, and revising employment packages. |
| Access | Spellbook offers a free academic access program for law students. Students can apply for access via the Spellbook website. |
| More Info |
Visit the Spellbook Learning Hub for learning resources. More information about Spellbook's features is available at the Spellbook archive on LawSites. |
| Tool | Vincent AI |
| Description |
Vincent AI combines vLex's global legal database with generative AI and workflow tools. It includes more than 20 "workflows" (e.g., "Ask a Research Question," "Compare Law in Different Jurisdictions," "Analyze a Contract") and supports custom workflow development. In Winter 2025, vLex introduced multimodal capabilities for analyzing audio and video content such as depositions, oral arguments, and client interviews. In November 2025, Clio acquired vLex in a transaction valued at approximately $1B. |
| Access | Vincent AI is not currently licensed for institutional use at UC Davis. A 3-day trial is offered at vlex.com/vincent-ai. Due to its pricing and enterprise-oriented licensing model, Vincent AI’s academic use is limited. |
| In Practice |
Vincent AI is increasingly discussed in legal technology contexts as part of the shift toward workflow-embedded AI in legal practice. A September 2024 LawSites article noted that Vincent AI integrates directly into firm workflows rather than functioning solely as a research engine. That same year, the American Association of Law Libraries (AALL) named Vincent AI its New Product of the Year. It was the first generative AI tool to receive this award. Clio's 2025 acquisition of vLex indicates further movement toward integration into a broader legal-work platform combining research, practice management, drafting, and automation. A 2025 randomized controlled trial assigned upper-level law students to complete six legal tasks using Vincent AI, an general-purpose reasoning model, or no AI. Students using Vincent AI were more productive in five of six tasks and encountered fewer hallucinations than those using the general-purpose model. |
| More Info | For additional information, visit the official product page. |
| Tool | VitalLaw AI |
| Description | VitalLaw AI is a premium generative AI tool available through Wolters Kluwer’s VitalLaw legal research platform. It supports research across multiple practice areas, including antitrust, cybersecurity and privacy, energy and environment, healthcare, intellectual property, securities, and tax. Built on Wolters Kluwer’s expert-authored legal content, the tool enables users to conduct conversational research, summarize documents, generate compliance checklists, and ask follow-up questions in natural language. |
| Access | VitalLaw AI is not currently licensed for institutional use at UC Davis Law. Students and faculty may explore individual subscription options via Wolters Kluwer’s website, which may include a free trial. |
| More Info | Wolters Kluwer provides several resources to preview and guide the use of VitalLaw AI. The Using VitalLaw AI help section explains how responses are generated and includes screenshots of sample outputs. The companion article Writing Effective AI Prompts offers introductory guidance and best practices. For additional product details, visit the VitalLaw AI product page or read the LawNext feature review. |
Custom-built tools use generative AI technology to serve a specific organization, such as a law firm, court, or legal clinic. These systems are often trained on private data and tailored to match internal workflows or document standards. For example, a firm might develop a tool that drafts contracts using its preferred structure, or a court might deploy an AI system to help manage case assignments. These tools are not typically available to the public and may be difficult to evaluate from the outside.
Athena, the proprietary AI assistant of American law firm Troutman Pepper, launched in August 2023 and built on OpenAI’s GPT technology. Deployed within the firm’s private cloud, it assists with tasks like drafting, summarizing, editing, and document analysis. By August 2024, Athena had over 1,000 active users, with features such as "chat with documents," image creation via DALL-E 3, and security and ethics safeguards, including mandatory training for users.
Law firm Ballard Spahr has launched three internal AI tools to improve efficiency and reduce costs:
The tools, locally hosted for security, aim to automate non-billable tasks and streamline workflows. Initially, the firm restricted AI use in client work but later developed structured policies for safer usage. These tools are part of the firm's Ballard360 legal tech suite.
DechertMind is a proprietary suite of generative AI tools developed by the global law firm Dechert and launched in April 2023. Initially introduced as a chatbot, it has since evolved into a more comprehensive platform for tasks like document drafting, review, and research. Built in-house by Dechert’s innovation team, DechertMind leverages custom AI models tailored to the firm’s specific needs, offering greater control and customization compared to third-party AI tools. In November 2024, Dechert was recognized by The American Lawyer for its innovative use of generative AI.
Davis Wright Tremaine (DWT) launched a generative AI chatbot in 2023 for firm employees, using the ChatGPT platform but restricting its information sources to the firm’s public-facing content. In 2025, DWT expanded its AI platform with DWT Prose, a tool that suggests writing edits based on the work of the firm's established attorneys. DWT Prose integrates multiple commercially available LLMs alongside custom AI models developed in-house. DWT also announced a partnership with Stanford University's CodeX to develop new AI-powered tools tailored to specific use cases identified by the firm.
fleetAI is a generative AI platform developed by Dentons, the world’s largest law firm by number of lawyers, operating in over 80 countries. Launched in August 2023 and built on OpenAI’s GPT-4 technology, fleetAI assists with tasks like document analysis, summarization, automated reporting, and clause extraction. It also supports multiple document uploads and generates context-based legal questions. Currently used firm-wide, fleetAI is set to expand globally and integrate with additional legal technologies, including future features like contract-specific automated reports.
Garfield.Law, the "first AI-driven law firm," was authorized by the Solicitors Regulation Authority (SRA) in England and Wales in May 2025. The firm uses a generative AI platform to help individuals and small businesses recover unpaid debts of up to £10,000 through the small claims court system. Their services include generating letters, forms, and other court documents. Regulated solicitors review all AI-produced outputs, and clients must approve each step in the process.
