The Pennsylvania State Capitol in Harrisburg. (Credit: Tom Sofield/LevittownNow.com)
In the wake of a Trump administration executive order striking state-level AI regulations, Pennsylvania will continue to prosecute its laws related to artificial intelligence in the interest of its residents, said state Attorney General Dave Sunday last week.
“The reality is, the executive order is kind of vague, for lack of a better term, and we’re going to have to kind of sit back and watch and wait and see what that actually turns into in Pennsylvania,” said Sunday.
Until a newly created federal task force took action, it would be difficult to know what the state’s response would be — though “obviously, we disagree with preempting state regulations,” Sunday said.
“If there’s a statute that is attacked for its constitutionality, then we will absolutely defend it,” he said.
As attorney general, Sunday would be tasked with defending Pennsylvania’s laws in court, including the bevy of new laws and pending legislation related to AI. Many proposals have focused on notification — such as a pending bill heard in a technology committee Monday — alongside criminal offenses, including those adding AI-generated images to the definition of child sexual abuse materials.
Sunday said he would “absolutely” recommend prosecutors continue filing those criminal charges, regardless of the executive order, related, for example, to the use of AI “deepfakes” to create pornography.
“We think it’s critical that we have the ability to prosecute people that use this tool for terrible, horrific things that hurt people,” said Sunday. “We have to be in a position to be able to hold those bad actors accountable.”
The executive order
President Donald Trump signed an executive order earlier this month overriding state laws, saying a national policy needed to be crafted. The new AI Litigation Task Force could bring a legal challenge against states with AI regulations, with exceptions for child safety protections and data center infrastructure.
In particular, Trump said AI companies needed one set of rules for approval or disapproval, rather than “50 different sources.”
Sunday said he could “understand … the thought behind that.”
“But, at the same time, we all have a job to do. I’m the Pennsylvania attorney general, and my job is to do everything I can to protect citizens, keep people safe, keep kids safe. So I will always argue for doing that in my job,” he said.
Earlier this month, Sunday led a bipartisan effort from 42 attorneys general to pressure leading AI companies to identify guardrails to reduce AI sycophancy — or “when an artificial intelligence model single-mindedly pursues human approval” — and delusions.
The Dec. 9 letter urges companies to respond by Jan. 16, citing real-world deaths (including murders and suicides) as well as hospitalizations for psychosis and “other delusional spirals” tied to AI use.
“AI can do amazing things as well, and we have to embrace that,” said Sunday. “That’s why we, as attorneys general, went directly to the companies to be able to … put us in a position where we hopefully, collaboratively, set these guardrails up.”
Sunday said the accompanying letter wasn’t penned “in reaction to” the anticipated executive order.
AI in the Capitol
State lawmakers, meanwhile, are considering a bill that would require health care providers and insurers to notify their customers about their use of AI, explicitly citing the lack of federal regulation as a motivator.
“Right now, there are no real federal regulations,” Rep. Arvind Venkat (D-Allegheny), the bill’s sponsor, told the House Communications and Technology Committee last week.
Specifically, Venkat worried about the “black box” of AI making clinical decisions that couldn’t be reproduced consistently, since the algorithms used are closely protected trade secrets.
Venkat’s House Bill 1925 got support from several stakeholders — including a representative from Sunday’s office, consumer advocates and providers — who pushed for human oversight.
An emergency physician, like Venkat, spoke about a possible benefit of AI. WellSpan Health’s David Vega, said that the technology presents an opportunity to “restore humanity in medicine,” citing the tendency of electronic medical health records to turn providers into “data clerks” during note taking.
“What I see as the successful implementation of AI … it has the promise to turn our focus back to the patients. To get us to stop looking at the computer screens and, instead, spend our time talking to patients,” said Vega.
Other providers noted that the bill extends reporting requirements to third-party contractors, even though providers might not know about their AI use.
The president and CEO of the Insurance Federation of Pennsylvania, Jonathan Greer, pointed to that electronic record precedence as a reason for federal intervention in AI. Prior to a national law, states had varying requirements that were difficult for insurers to meet, he said.
“If individual states are allowed to do it themselves, you will create friction,” Greer said. “It is made for a federal fix.”
Committee chair Rep. Joe Ciresi (D-Montgomery) closed the meeting without a vote, but concluded by saying that “Congress has fallen behind … and it’s only going to get worse.”
“I would hope that the federal government would move but sometimes the states need to move first and make the federal government wake up and do something. Because, right now, there is a lot of willy-nilly that exists,” said Ciresi. “The federal government is lollygagging.”
