Gunderson Dettmer Launches ChatGD: First U.S. Law Firm to Develop Proprietary Internal Generative AI App

Silicon Valley-based international law firm, Gunderson Dettmer, has launched an artificial intelligence application known as a generative AI chat app, named ChatGD. The firm, which boasts more than 400 lawyers and has numerous clients in the artificial intelligence field, holds the distinction of being the first U.S.-based firm to create a proprietary internal tool using generative AI technology.

Just last week, global law firm Dentons announced its intent to roll out a proprietary version of ChatGPT, though a definitive launch date is yet to be revealed. ChatGD is Gunderson’s first internal AI tool—a development the firm hails as a significant landmark in harmonizing its internal legal expertise with technological and engineering advancements.

In its current use, some of the firm’s attorneys use ChatGD to query and alter documents using a secure, enterprise implementation of OpenAI’s models via Microsoft Azure, ensuring the safe handling of confidential data. The firm views AI as an accelerant to its work, not as a replacement for human judgment or expertise.

ChatGD’s framework is model-agnostic, meaning it can engage with other language models like Anthropic, Google Palm, and Meta’s Llama 2. “We think it’s very important not to get tied to a single vendor or AI model at the moment,” explained John Scrudato, the firm’s legal engineering and data strategy manager. He added that new models from companies like Google and Meta continue to emerge, highlighting the importance of building a flexible, adaptable tool.

ChatGD consists of two main components: a general chat mode, and the more intriguing retrieval-augmented generation (RAG)—a method using salient external data to enrich queries. Lawyers can upload documents or groups of documents, then pose queries to the language model which returns responses based on the context provided by the documents. This not only allows lawyers to question the model based on their own internal knowledge, but it also minimizes inaccuracies.

Further outlining the tool’s ability to support daily work, Scrudato gave an example of how a lawyer could modify an indemnification clause that risks more to the investor by simply uploading a model financing agreement to ChatGD and asking it for a newer version of the clause. The same method could be used to streamline responses to recurring requests by uploading documents the firm has already created for specific tasks and generate new versions to answer new requests, eliminating much of the recurrent work.

Looking at the future, Green and Scrudato are already envisioning continuous development of ChatGD. One immediate objective is to connect the app with the firm’s internal knowledge repositories to enable users to chat with particular documents uploaded by individuals, as well as document collections created by the firm. They are equally exploring options to connect the app to the firm’s complete document management system.

Originally published on LawNext.