LegalTech Diaries Volume 19
Cecilia Ziniti
Co-Founder & CEO
GC AI
LegalTech Diaries Volume 19
Cecilia Ziniti
Co-Founder & CEO
GC AI
Your career runs from a paralegal role at Yahoo in the early internet days through Alexa at Amazon and three General Counsel roles. Which of those chapters shaped GC AI most directly?
Alexa set the direction, and the GC roles shaped the product itself.
I started on Alexa in 2013, when the device was still in alpha and beta. What I took from that was that AI would affect every part of our lives, and I made a career bet on it. After Amazon I worked on legal issues at a robotics company and then at Cruise on autonomous vehicles. AI stayed at the centre of my career, with a steady mix of IP, privacy and consumer protection issues.
My three GC roles guided what GC AI does. In-house work runs on company-specific playbooks, cross-functional stakeholders and a risk tolerance that changes from deal to deal. Most legal AI was built for law firms, or it was a general-purpose tool that happened to have a legal use case. I built for the job I had actually done.
Yahoo was where it started. I came to Silicon Valley in 2002, while Yahoo was leading in search and sponsored search, and watched the internet become a business. Since then I’ve been lucky to be a lawyer through every wave: mobile, then cloud, and now AI.
You’ve described GC AI as the product you kept wishing existed. Was there a specific moment as a GC when you stopped wishing and decided to build it yourself?
Yes, and I remember the feeling well. In early 2022, before ChatGPT launched, I got access to early language models. I used one to draft contract language and review an agreement, and it gave me the chills. I was at Replit at the time. Replit had a software engineering AI, and I was watching our engineers use AI every day, so I started using it for my legal work.
It felt like the first time people used an iPhone. Earlier legal tech had helped, but it stayed out of the core work: the drafting and the analysis. This went straight into that work.
The idea didn’t become a company overnight. My co-founder Bardia and I started with a newsletter about AI and legal, and then we began teaching lawyers how to prompt. Those conversations with early adopters showed us what in-house teams needed. On 1 November 2023, we committed full-time, and General Counsel AI, Inc. was incorporated a week later.
You’ve argued that general-purpose tools fall short for professional legal work because of citation, context and privilege. Which of those gaps is hardest to close, and how does GC AI approach it?
Context is the hardest. Citation and privilege are hard too, but they are mainly engineering and contracting problems with clear answers. Context depends on the judgement of the person asking.
On citation, GC AI gives source citations with hover previews and highlights the cited text in the document. A lawyer can check every answer in seconds, which is how lawyers should work anyway.
On privilege and confidentiality, customer data is never used to train the underlying AI models. We add encryption and data isolation at the organisation level on top of that. No tool can promise that privilege is preserved. What we can do is make sure the tool doesn’t create the exposure.
Context is harder because the same indemnity clause can be fine in one deal and unacceptable in the next. When a team signs up, we onboard GC AI the way you’d onboard a new hire. It learns the company’s industry, jurisdictions, risk tolerance and preferred clauses. From there, it shapes its answers to the company’s legal positions, regulatory environment and standard practices. Playbooks then capture the team’s positions so the review reflects how that team actually negotiates.
