How Colorado is leading insurance AI regulation

In 2021, Colorado Governor Jared Polis signed SB21-169, a first of its kind law that would establish a framework to require insurers to test whether the use of big data in their processes is discriminatory. The first draft set, Algorithm and Predictive Model Governance Regulations was released in February and focuses on life insurance underwriting. Next up is private passenger auto underwriting, and The Colorado Division of Insurance, part of the Department of Regulatory Agencies in the state, is working to draft various rules for insurers. 

Michael Conway, the Colorado Insurance Commissioner, said during the April 6 meeting, that this work has to be done.

"The big data/AI world is obviously coming at us very fast, in new ways, almost on a constant basis," Conway said. "I absolutely, fundamentally, believe that if we as regulators and industry don't figure out how to regulate it within the insurance practice that somebody from outside the insurance world is going to come in and regulate it for us, and I don't think that will be a good thing for anybody.

"I really do hope that you all will engage with us in good and candid conversation, but know that we're going to continue to push this work forward," he continued.

The stakeholder meetings are meant to engage people within the industry in a conversation prior to the adoption of rules. The law, Restrict Insurers' Use of External Consumer Data, protects consumers from unfair discrimination on the basis of race, color, national or ethnic origin, religion, sex, sexual orientation, disability, gender identity or gender expression.

Jason Lapham, big data and AI policy director for the Colorado Division of Insurance said that while the first segment on life insurance is still in process, it's time for the auto insurance industry to prepare as well.

"The first order requirement for carriers is that they must test whether the use of those tools that use external consumer data and information sources result in unfairly discriminatory outcomes, as well as establish a governance or risk management framework that is meant to mitigate against potential consumer harm that can result from the use of those tools," he said.

External consumer data is defined as credit scores, social media habits, locations, purchasing habits, homeownership and more, Lapham added.

"One example is the use of credit histories [which] are often a reflection of historical economic discrimination against historically marginalized communities and the use of insurance scores that include credit histories, to assess risk, have the real possibility of reproducing that discrimination by disproportionately charging members of protected classes higher premiums," Lapham said.

Conway said that the division doesn't know what kind of external consumer data each individual company is using, there likely are companies that are using criminal history, for example.

"What we're going to be focusing on when we're talking about the testing is the overarching outcomes of whatever the algorithm or big data mechanism is, that the insurance companies are using for underwriting and then we'll work backwards from there to see what rating factors we think or the industry believes are having the potential negative impact, unfairly discriminatory impact. That is a long way down the road," Conway added.

Lapham said there are potential benefits of using algorithms that include big data and predictive models including carriers being able to better assess risk, which could create opportunities for insurers to provide coverage to underserved populations.

Lynne Elliot with the APCIA, spoke during the meeting to note that the APCIA represents about 1,200 insurers and reinsurers in the Colorado area that provides 57% of the property and casualty coverage in the state. "Senate Bill 169 creates unique challenges for regulators and regulated companies," she said.

Elliot also highlighted the APCIA and other industry efforts to increase diversity, equity and inclusion.

Michael Dylon, with the Consumer Federation of America, added: "We appreciate that insurance industry's commitment to diversity, equity inclusion, but that shouldn't only be limited to within your workforces. You should also look at diversity, equity inclusion, as it relates to your business practices and the use of the various factors that we talked about that he listed, including credit history, homeownership, educational attainment, occupation, and others. And we have abundant evidence that the use of these factors disproportionately hurts Black and Latino consumers, and that hurts other consumers as well, unfairly."

The National Association of Insurance Commissioners' Innovation, Cybersecurity, and Technology (H) Committee has several groups working to identify and develop a framework for the use of artificial intelligence. The Accelerated Underwriting Working Group will examine the use of external data and analytics in accelerated underwriting in life insurance, which works closely with the Big Data and Artificial Intelligence Working Group. That working group is charged with looking into the use of big data and AI, as well as ML, for development of potential recommendations for model governance. 

The European Union is also working on a law related to AI. The legal initiatives include a European legal framework, a civil liability framework and a revision of sectoral safety legislation, according to the EU website.