While Congress continues to work on putting guardrails around artificial intelligence, states are developing standards and rules for companies that use the rapidly evolving technology.
Several states, including California, Illinois, Texas, and Colorado, have introduced or passed laws focused on protecting consumers from harm caused by AI. Companies, insurers and government agencies have long used AI to make hiring, lending and housing decisions, but there is widespread concern that the technology will make biased and discriminatory decisions.
In fact, research shows that algorithms and automated decision-making tools can lead to discriminatory decisions.2020 study published in Journal of General Medicine We have shown that AI algorithms reduce the chances of transplant referrals for black patients. Another 2016 ProPublica report found that AI software used to conduct criminal risk assessments in Broward County, Florida, falsely identified black defendants as likely to commit future crimes. revealed that it was probable.
Congress has yet to move forward with a bill regulating AI, and is now grappling with new technology challenges brought on by the growing popularity of OpenAI’s ChatGPT. Policymakers have held several public hearings this year to better understand the risks of these technologies, from piracy to the spread of AI-generated misinformation.
In a similar pattern to how states are addressing data privacy without overarching guidance in federal data privacy laws, each state is addressing AI concerns with its own rules as Congress stalls.
countries take action
Colorado is honing rules for predictive algorithms used by insurance companies, which could pave the way for future regulation of AI use in other industries.
In 2021, the Colorado Legislature will authorize external consumer data, such as educational background, credit scores, and social media and online purchasing habits, from insurers, as well as algorithms and programs that use such data to unfairly discriminate based on race. Passed a bill restricting the use of predictive models. , religion, gender and other factors. The Colorado Department of Insurance, tasked with adopting rules for certain types of insurance, released its first set of draft rules for life insurance companies in February. The draft rule establishes requirements for the governance and risk management framework of predictive models used by insurance agents.
[Colorado is] It was really expected to be the spearhead of national regulation on this issue. Not only in the insurance industry, but in other fields as well. Mary Jane Wilson BilikPartner, Evershed Sutherland
The state has allowed public comment on the draft rule, and companies are now waiting for a final proposal, said Mary Jane Wilson-Billik, a partner at law firm Evershed Sutherland. The state also plans to release another draft rule for bias testing in AI systems later this year, which Wilson Bilik said will be important.
How to test AI systems for prejudice and discrimination is a problem regulators outside of Colorado have struggled to understand, and even regulators in the European Union are struggling to set standards for AI legislation. there is
“They are really expected to lead the regulation on this issue nationally,” she said of Colorado. “Not just in the insurance industry, but in other areas as well.”
Meanwhile, California legislator Rebecca Bauer-Kahan introduced a bill earlier this year to regulate automated decision-making tools. The bill would require those who use automated decision-making tools to notify individuals of their use and would ban tools that cause algorithmic discrimination. Similarly, city councilors in Washington, D.C. introduced a bill requiring companies to evaluate the AI algorithms they use for signs of bias.
When Congress Acts
Wilson Bilik said he believes Congress recognizes the potential harm, as shown by the blueprints for NIST’s AI Risk Management Framework and the White House’s AI Bill of Rights. But regulators are trying to think through how to regulate innovation without killing it, he said.
“The same principle is expressed in different ways,” she said. “From now on, the question is whether we will put all this into practice.”
Along with issuing guidance on AI adoption, several enforcement agencies, including the Federal Trade Commission, have announced that existing consumer protection laws will apply to companies using AI systems until new AI regulations are proposed. , which has been demanded by several US Senators. The U.S. Department of Commerce also recently issued a request for information on AI liability actions.
In fact, AI regulation is a hot topic, says Cameron Kelly, a global thought leader on AI and visiting fellow at the Brookings Institution.
“There’s a lot of interest in how to put guardrails around, in terms of accountability, transparency, risk assessment, performance and results measurement,” he said.
However, Kelly said he wouldn’t be surprised to see states addressing the issue in the meantime, although a federal AI rule is a possibility, and that both federal and state governments will continue to make AI rules. He said he expected to do so.
“I think we’re going to see both,” he said. “While I believe in the need for strong and consistent national regulation, there are aspects of this that have historically been state regulation. I did.”
Makenzie Holland is a news writer covering major technology and federal regulations. Prior to joining TechTarget Editorial, she was a general reporter. Wilmington Star News crime and education reporter wabash plain dealer.
