Exploring CaseMark's Pioneering Approach to AI Prompting in the Legal Sector.
Earlier this week, the US Patent Office published its guidance around AI and how it relates to inventorship and incentivizing new IP as it relates to artificial intelligence. Buried beneath the fold of that post was a confirmation for something we’ve known all along; contributions, specifically in the form of prompting, can be potentially considered IP.
When we started CaseMark 6+ months ago, we saw the power of prompting early on. So much so that we actually created a step-by-step guide to prompting for legal professionals. Legal Prompt Guide has gone on to help thousands of legal professionals increase their prompting acuity as it relates to AI and the law.
While we have made some forays into fine-tuning for LLMs, we found the pejorative “juice wasn’t worth the squeeze”. It’s hard to find decent data sets of meaningful size to be able to fine-tune with. The results were lackluster at best. Instead, we doubled down on prompting and it has paid off.
Our approach to prompting and how we’re solving problems for our customers is the cornerstone of what makes CaseMark. Not only that, the prompting approach allows us to stay agnostic when it comes to which technologies we use on the LLM side of things. This gives us flexibility and allows us to future proof our platform and take advantage of the investments being made into AI infrastructure by the biggest players in the market.
One major concern we hear over and over from our customers is about the security and privacy of their sensitive client data. While most AI companies out there today simply wrap OpenAI’s GPT4 API, we here at CaseMark have taken a more robust approach. In addition to partnering with Microsoft Azure, we’re able to deploy standalone instances of GPT4 and other LLM instances that we can then load and query our customers’ data against, all while siloed off from the outside world. There just is not a way for sensitive client data to leak to 3rd party services and we will always stick to our promise never to train with customer data.
Being a legal-first AI company means we’re building a solution that can be trusted by legal professionals. It’s the same reason we’re in the midst of our SOC2 Type II certification so that we can point to a 3rd party that is regularly verifying our capabilities. Doing this is just table stakes for the legal space.
When thinking about how all of this relates to prompting and our customers’ processes, we will soon be taking it a step further and we are excited to see the USPTO in agreement with us on how this will all relate to defensible IP. Keep an eye on this space for more coming soon.
--
About CaseMark
At CaseMark, we're pioneering the integration of artificial intelligence in the legal sector. Our suite of tools, including Workflow and Productivity are designed to enhance the efficiency and effectiveness of legal professionals. Leveraging the power of advanced AI models in a secure and user-friendly platform, we provide solutions that are not just innovative but also respect the integrity and ethics of the legal profession.
Connect with Us
Follow Us
CaseMark – Empowering Legal Precision: One Workflow at a Time.
