The compliance demands of the rapidly evolving artificial intelligence governance landscape are creating mounting work for enterprises left to manually review and verify the legality of certain outputs of their AI systems.
To help organizations automate risk assessments and check their compliance posture with regulations, Luminos.AI and ZwillGen have partnered to launch the ZwillGen Law Library as a new feature that is now available for Luminos platform customers and ZwillGen clients.
The ZwillGen Law Library will allow customers to automate AI risk assessments for predictive models and test their models for issues, such as bias and deidentification risk, while leveraging the AI expertise of ZwillGen and Luminos. The platform features detailed compliance questions for the relevant organizational team to answer for the given AI use case, in addition to model testing through Luminos.
ZwillGen AI Division Director Brenda Leong, AIGP, CIPP/US, said in an interview with the IAPP that the new law library was developed for eight common issues impacting AI governance practices. They cover testing capabilities, such as deidentification and bias risk, and regulatory compliance in the context of rules in the EU and Colorado.
Leong said, once customers begin using the platform and run into new AI-related problems, she wants them to raise the issue with Luminos.AI and ZwillGen so the new issues can be automated and incorporated into the law library as well.
"The more AI a company is using, the more governance burden they have," Leong said. "Both the more official explicit legal requirements, as well as the soft regulatory governance requirements they have, which is all difficult without a significant commitment of resources in terms of expert people, outside legal services and internal technical capabilities.
"So we asked ourselves how can we fundamentally give our clients better service on some of the things that they need to do that we’re giving them legal advice for, but is either too overwhelming or too expensive for them at scale?"
Today, businesses are adopting any number of hundreds of AI models for various business applications, according to Luminos.AI CEO Andrew Burt. He said just five years ago while working at Luminos.Law — which was acquired by ZwillGen earlier this year to be its in-house AI arm — a regular part of his work was consulting with clients and reviewing each model they put into use for a compliance check.
Since then, the uptick in AI adoption has necessitated the need for creating an automated means for companies to conduct a significant portion of their risk assessments without engaging in-house or external legal counsel to review every model the company wants to use.
"On one hand our customers needed very high-end service and very specific legal advice, then on the other hand, for their other AI models, they still had issues that were not as nuanced or acute, they needed something they could automate. No one had built software capable of that, which is why we started Luminos.AI," Burt told the IAPP. "Without a solution like this, there's no way businesses are going to be able to adopt AI at scale because they're going to need to be able to use thousands of AI models that need to be reviewed and approved for legal risk."
The goal of creating the ZwillGen Law Library is to help businesses identify what aspects of their AI governance programs are repeatable and automate those processes. That way the organizational stakeholders overseeing AI governance efforts can keep their focus on their high-risk outputs.
"Customers should ask themselves what is replicable, because if it's replicable, they should be using software," Burt said. "For whatever is not replicable, they should be going to lawyers and people who can sit down and do the careful, high-end analysis that's required."
Leong echoed that automated risk assessments produced through the platform are not a substitute for a lawyer verifying and signing off on the assessments' respective veracity. However, she said the outputs using the law library can then be used by an organization to engage in-house or external counsel to get sign off or explore unresolved issues after conducting the assessments.
"There are limits to what can be automated, so we are not saying that whatever (customers) answer those questions with guarantees that they are the compliance standard," Leong said. "We are saying that if you follow the form and this process via the platform operated by Luminos.AI that it is asking you the right questions and guiding down the right path to put you in that position to comply."
Alex LaCasse is a staff writer for the IAPP.