Beyond legal: exploring AI & climate through a legal lens

AI developments, such as Anthropic’s new legal tools for Claude, are disrupting financial markets and prompting lawyers and executives to ask how AI will affect the practice of law.

At The Chancery Lane Project (TCLP), we don’t think that’s the most interesting question. Instead, we want to know, in an era where anyone has access to the means to produce legal-looking content, what does good law actually look like?

We’ve been answering that question since we were founded. Our library of Clauses, Glossary Terms, Guides and Playbooks is the high water line of climate-aligned contracting. But we need to stay relevant in a world where making agreements, contracts and conducting business is done through, or at least alongside, AI & large language models. 

We’re actively exploring how TCLP can support the intersection of climate law and legal tech, drawing not only on our existing library of legal content but also on the expertise of our brilliant subject-matter specialists and the organisation’s deep tacit knowledge. With support from the Patrick J McGovern Foundation, we are sharpening and accelerating this work. In partnership with Unboxed, we are building practical prototypes designed to solve real-world problems and iterating quickly so we can share our progress openly.

This post is the first in a series exploring this space.

Our process
We’ve created internal tools to help us understand and annotate the legal concepts and entities that sit in our content. These will allow us to ingest and understand legal content at scale and extend the graph of our knowledge to encompass sources like Climate Policy Radar‘s incredible database of global climate law. We’ve layered semantic information on top of our published content so that our own website acts as the foundation of the knowledge graph and gives semantic context to AI tools visiting our site. This means we can begin to act as a canonical legal infrastructure for the tools that legal tech and the big AI firms are building. 

We’re also building tools to mark our own homework. We haven’t allowed an LLM to annotate our content and accepted a certain amount of noise in that system. Instead, we brought together our in-house experts and friends in top-tier law firms to help us get this right. Similarly, we haven’t immediately published our semantic layering because we wanted to ensure that when we do, we would be able to quantify the improvement that it made. It would be easy to vibe code a few interfaces that are nothing more than well-styled wrappers over a simple prompt to Claude, but we want to build real foundational tools.

We’re also building a number of ‘provocatypes’ – lightweight prototypes to spur conversation and help us understand which tools lawyers and their professional collaborators would actually use. They also allow us to explore how to create demand for more planet-positive legal content and to evaluate the impact that existing contracts and legal documents might have. It’s crucial for us to be driven by user research in understanding where to deploy our limited resources in building tools to effect change.

There are risks in doing it this way. We might end up moving too slowly, either out of caution or a simple lack of resources. Legal tech firms might care more about greenwashing than embedding planet-positive legal thinking in their tools, or we might just not keep up with the pace of change.

While we strongly believe that applying these tools correctly now now will save vastly more carbon emissions in the future, we are also mindful that using them creates carbon emissions and other environmental impacts. We plan to discuss in more detail how we ethically justify our use of AI tools. Until the large-scale providers give more insight into how much carbon their applications use, we are going to be as smart and sustainable as we can be in our use, and are actively looking for transparent and lower-carbon partners for this tooling.