Two FiscalNote alums founded a startup for AI accountability

Despite AI increasingly permeating everyday technology and life, founder Andrew Gamino-Cheong sees an enduring flaw in its development.

“If I wanted to come up with a new flavor of potato chip, there’s a lot more regulation I have to go through that if I wanted to come out with an AI system that could decide how long somebody goes to jail for — and that seems intuitively wrong,” Gamino-Cheong told Technical.ly.

The 2021 RealLIST Engineer doesn’t think many people fully understood AI before ChatGPT came along and brought a concrete example of its capabilities before the public. But with the growth in awareness, he and cofounder Gerald Kierce want to show companies how to how AI accountable — and how to use it responsibly.

The two alums of political software firm FiscalNote created Rosslyn, Virginia-based Trustible AI to help companies stay compliant with AI legislation while building responsible tech. Founded in January of this year, the company just emerged from stealth with a $1.6 million pre-seed round led by Harlem Capital. VamosVenture also took part in the round, as did angel investors Julius Genachowski; FiscalNote cofounders Tim Hwang and Gerald Yao; Jim Hunt; Joe Zhao; and Neal Parikh.

Trustible, as described by Kierce, helps companies build trust and manage risks from AI technology. The SaaS platform helps companies answer questions like why a piece of AI tech was built, how the data is collected, who’s responsible for that data, who’s participating in a project and how stakeholders’ input factors in. It simultaneously tracks regulatory compliance and relevant laws. In other words, Kierce said, the company brings the principle of “say it, do it, prove it” into AI development.

“We want to help guide organizations on: Here’s the actual bias or fairness testing you should do based on your use case; then we can deploy different examples and structures of how to do that inside of your own ML environments,” Gamino-Cheong said. “There are actually resources for how to test for bias. There’s not a good resource about how to know what bias to test for.”

The four-person company is building the cloud-based platform in the hopes of making it easily deployable inside cloud environments. Kierce said Trustible was also incorporated as a public benefit corporation to ensure the company was growing and building technology in an accountable way, as well.

This tech is necessary, Gamino-Cheong said, because AI building methods are different. Now, tech can be developed with a few lines of code and API call-out from Google or Amazon. He added that this means teams aren’t as aware of their tech’s risks since it’s not built from scratch, he said.

“It’s no longer a team of PhDs in AI building your systems; it could be anyone and everyone,” Gamino-Cheong said. “That comes with its own set of risks and challenges that’ll be very complex for an organization to manage.”

To help, Trustible offers workflows and checklists as well as best practices when companies are customizing technology for their organization.

The new funds are a sigh of relief for the founders, who were part of the local founder community impacted by Silicon Valley Bank’s collapse (they’ve since moved to Bank of America). They’ll spend the funds on building out the products, hiring and finding pilot customers. They also plan to open a seed round “fairly soon.”

As it grows, though, Kierce said that the company remains dedicated to remaining in the DMV. Trustible currently operates in a hybrid workspace model in Rosslyn because the founders feel that early-stage collaboration is important in a company’s initial stages.

“We’ve benefited greatly from the DC tech ecosystem at FiscalNote,” Kierce said. “We, in many ways, want to be able to contribute back to the tech ecosystem and participate in it.”