As the marvels and shortcomings of artificial intelligence increasingly capture headlines, Pennsylvania lawmakers are proposing legislation to track and regulate the emerging technology’s use.
AI – technology that mimics human intelligence to perform tasks – has emerged in the last year as a transformative force in areas ranging from art to business decision making.
Its critics and even its creators warn of the potential for AI to cause harm through misuse, faulty decisions or unintended outcomes that could pose an existential risk to humanity.
A slate of bills introduced by Democratic state Reps. Chris Pielli, of Chester County, and Robert Merski, of Erie County, would set Pennsylvania on the first steps toward better understanding AI and guarding against its negative impacts.
“This is one of the most important issues in my opinion facing mankind,” Pielli told the Capital-Star, adding that he sees the issues as a new iteration of the age-old problem of new technology giving rise to greed at the expense of the people whose lives it affects.
“There is a quick profit to be made by using these programs but we need to look at what is going to be the long-term cost,” Pielli said.
Merski and Pielli separately have proposed legislation to study the development and uses of AI and how it impacts the commonwealth and its residents.
Merski’s legislation tracks a companion bill in the state Senate proposed by Sen. John Kane, D-Delaware, that would direct the Joint State Government Commission to create an advisory committee on AI.
Pielli’s bill would create a taskforce to study whether there is a need for a state agency to monitor and license AI products used in Pennsylvania.
Warning that text generated by AI is often riddled with inaccuracies or invented “facts” that researchers call hallucinations, Pielli proposed legislation that would require AI-generated content to carry disclosures to guard against the dissemination of false information.
While AI has the potential to improve state government services and save taxpayers money, Pielli said the ways in which the commonwealth employs the technology should be disclosed and tracked. Pielli proposed a bill that would create policies for the development, procurement, implementation and ongoing assessment of systems using AI.
Merski and Pielli also have jointly proposed legislation that would make a criminal offense for the unauthorized dissemination of AI-generated “deep fakes” – in which a computer is used to create realistic video or audio impersonations of people.
Fewer than a dozen states have enacted legislation that regulates AI, but Pennsylvania is not alone in analyzing how it should approach the technology, Heather Morton, director of financial services, technology and communications for the National Conference of State Legislatures, said.
About 10 states have launched a study or task force to examine how AI technologies might affect their workforces and regulate how it is used by government agencies and private companies, Morton said.
Many pieces of proposed legislation focuses on the responsible use of AI, such as a bill in the New Jersey Assembly that would regulate the use of automated tools in hiring decisions. The goal of the legislation is to ensure that AI algorithms don’t incorporate biases that would lead to discrimination. California and New York are among states with similar proposals, Morton said.
”Legislators want to make sure this evolving technology is complying with [anti-discrimination] laws and incorporating those principles,” Morton said.
Other new laws proposed by state lawmakers concern aspects of the technology affecting consumer protection, data privacy and the deployment of AI tools in the delivery of state and local government services.
The greatest challenge for lawmakers is balancing the need for regulation with the desire to foster development of new technologies, Morton said.
“There are fears that it will eliminate jobs,” Morton said. “But on the other hand it may create new jobs.”
University of Pennsylvania Law School professor Cary Coglianese said lawmakers setting out to understand and regulate AI need to recognize that it is not a single monolithic entity.
“It is a variety of different analytic tools that are applied in a variety of different applications,” Coglianese said. “The use of AI is something that can be found in almost any part of the economy.”
Because it is so varied and evolving rapidly, AI is not an easy thing to regulate. So the best approach for states is not to regulate AI itself, but how the technology is employed in industries over which the state government has primary regulatory authority, such as banking, insurance and utilities, Coglianese said.
He said the best approach is to implement regulations that require anyone employing AI to engage in stringent risk management practices that minimize the possibility of a catastrophic outcome.
“We can specify a goal but as humans we don’t always fully specify a goal and the AI tool may pursue the goal in ways that are misaligned with our full objectives,” Coglianese said, noting that the chances of a existential threat materializing are vanishingly small but the consequences are too great to ignore.
Our stories may be republished online or in print under Creative Commons license CC BY-NC-ND 4.0. We ask that you edit only for style or to shorten, provide proper attribution and link to our web site. Please see our republishing guidelines for use of photos and graphics.