In a groundbreaking moment, AI successfully negotiated a contract with…AI.
There were no real stakes involved, since it was a just a live demo, but artificial intelligence company Luminance just provided a glimpse of what the future might look like for the legal industry. On Tuesday, the UK-based company’s large language model (LLM) automated a contract negotiation “without human intervention, between two opposing parties.” Luminance claims this is the first completely AI-powered contract negotiation. The model was trained on 150 million legal documents to gain legal knowledge.
Industry-specific LLMs are one of the latest evolutions from generalist models like ChatGPT. Expect to see more of these customized AI models cropping up.
On Monday, OpenAI announced custom GPTs, which allow the user to build and train a model for their own purposes without any coding experience needed. GPTs are marketed towards ChatGPT users, but also ChatGPT Enterprise customers who use a private version of ChatGPT internally.
The benefit of using a bespoke model is not only data privacy, but also focused expertise for companies looking to automate tasks. Luminance, which does this for legal contracts, is aiming to cut down on hours lawyers spend negotiating terms.
Reviewing legal documents is a time-consuming process for lawyers. LLMs like Luminance’s proprietary legal-focused model, Autopilot, are able to understand and analyze massive amounts of information in a matter of seconds.
Automating the review of routine contracts like nondisclosure agreements could save legal professionals tons of time. “By putting the day-to-day negotiations in the hands of an AI that is legally trained and understands your business, we’re freeing lawyers up to focus their creativity where it counts,” said Jaeger Glucina, Luminance chief of staff, in the announcement.
This doesn’t mean there’s zero oversight. According to CBNC’s Ryan Browne, who saw the live demo, there’s a live log highlighting changes to clauses and suggested changes, so lawyers can review the process and double-check it for errors.
Topics
Artificial Intelligence