Using artificial intelligence to help inform decisions rather than to make decisions in place of people is the best way to comply with a patchwork of state laws surrounding the use of data and artificial intelligence, a panel said at JPMorgan Chase's Optimization and the Path to Innovation event Wednesday in Philadelphia.

For artificial intelligence and machine learning technology to properly work, there needs to be a large amount of diverse data plugged into it. However, the data plugged into the machines may be biased and new data privacy laws coming from the different states are beginning to impact how artificial intelligence is used.

"The biggest challenge is staying on top of what is happening in the U.S.," Tess Blair, a partner at Morgan, Lewis & Bockius in Philadelphia, said.

While the California Consumer Privacy Act has been getting a lot of attention, Blair explained bills governing data privacy are being considered in Nevada, Illinois, Maine and Washington. Earlier this year Illinois' Artificial Intelligence Video Interview Act went into effect. The law requires employers to inform prospective employees if artificial intelligence will be used during the interview process. Legislators in New York City and in New York state have introduced similar bills.

"We're down to that level of granularity here," Blair said.

Some companies have been using artificial intelligence to score prospective candidates during interviews based on facial expression and tone. However, critics of that use of artificial intelligence have argued that the machines are biased.

"There are examples that I've heard of where there is clearly bias built into these things," said Marc Lederman, co-founder and general partner of NewSpring Capital in Radnor, Pennsylvania. "It is unintentional bias in many cases because people built these programs based on their view of the world."

Outside of the hiring processes, the panel said artificial intelligence technology should be used to supplement information. Blair said companies are at a greater risk of running afoul of regulators if they rely too heavily on machine learning.

"Until we work through a lot of these things, it has to be that we use a model that informs the human being who has the final say," Anish Bhimani, chief information officer for commercial banking at JPMorgan Chase, said.

Before beginning to use data for artificial intelligence purposes, Blair told the crowd that companies and their counsel should be looking at the life cycle of data they collect in three ways: data acquisition, security and sharing.

"Depending on what industry you're in, you need to have all of that in place before you do anything with the data," Blair said.