Virginia lawmakers weigh risks and rewards of AI chatbots

This article was originally published by Virginia Mercury.

As more Virginians turn to artificial intelligence for simple tasks or even companionship, lawmakers are exploring policy ideas to address how AI chatbots are influencing residents’ health and well-being, as well as their privacy and security.

But policymakers face a complex web of obstacles in crafting regulations, given the challenges of the technology and broader legal, political and social environments.

Chatbots have proven both helpful and risky for users.

“It’s the very versatility and accessibility of these AI chatbots that make them both a really exciting technology, a very usable technology, but also present some very real risks to users,” said Kira Allmann, chief policy analyst for the Joint Commission on Technology and Science (JCOTS), on Monday.

That day, Allmann provided an overview of chatbots to JCOTS, covering types of bots and policy issues.With the assistance of artificial intelligence, chatbots communicate in a user’s language to help complete tasks such as searching the internet or creating content.

However, experts have identified significant policy concerns with their widespread adoption. In some cases, users have developed unintended emotional connections with chatbots and may share private information with them.

Developers have also created chatbots that can be used for malicious purposes, such as stealing personal information, and chatbots themselves are vulnerable to data leaks. Other concerns include how they determine their answers and their ability to influence society. 

“It seems like this is going to be really challenging to find legislation to put guardrails around this,” said Del. Anne Ferrell Tata, R-Virginia Beach. “It sounds like even when they know, they still are interacting and it just goes right over their head.”

In other states, Allman noted, policymakers have introduced or passed legislation requiring chatbot disclosures and age-appropriate design features.

Commission staff identified three types of bots, each with different ways of interacting with users. AI assistants help with a range of tasks, while AI companions are designed to provide emotional support or social interaction. AI agents, by contrast, are built to understand their surroundings, process information, and take actions based on either programmed directions or their experiences.

Allman noted that as more users turn to chatbots for mental health support — and as they are increasingly marketed as emotional support tools — no regulations currently exist in the healthcare system.

“They haven’t been vetted by clinical professionals,” Allman said. “They don’t necessarily follow any particular kind of rules or regulations associated with what we might consider the health space. They really are just technological tools, but people are leaning on them heavily as mental health support.”

Other areas the group discussed included protecting vulnerable populations seeking mental health support from chatbots, addressing implicit and explicit biases, and focusing on the dynamics and outcomes of chatbot use rather than just the technology itself.

Del. Cliff Hayes, D-Chesapeake, commission chair, said while he looks forward to continuing the conversation but emphasized that the overview served as a reminder for lawmakers to always consider the human element in AI interactions.

“This might be a reminder to us all that as we’re dealing with this technology that we always, always, always keep humans in the loop,” Hayes said. “That’s going to be very important when you’re talking about spaces that govern. But then, what do you do when there’s no governance in place and an individual is actually interacting?”

VCCS Chatbot Efforts

Virginia’s Community Colleges (VCCS) have used chatbot technology for more than a year, assisting students with a wide range of inquiries, from financial aid to enrollment dates.

In a May 2024 announcement, VCCS noted that providing around-the-clock student support can be “challenging undertaking, especially when budgets are stressed and resources are limited.” With the chatbot handling routine questions, staff have been able to focus on more complex issues.

At Tidewater Community College, email volume dropped by 60% and calls by 30% after the chatbot launched, according to Holly Whistler, coordinator for the Online Virginia Network.

“This reduction happened because users found the answers they needed, and staff was able to address other pressing and complex issues much sooner,” Whistler said in a statement.

VCCS said last spring that the chatbot — developed by Intrasee — interacts with users at least 500 times a day. When the bot cannot provide an answer, users can still leave an email or make a phone call.

Government’s AI Efforts

According to the Virginia Information Technologies Agency, artificial intelligence is already used by many state agencies to “process data, produce automated decisions, enhance customer services and increase government efficiency.” 

Virginia is also deploying AI to support its education, workforce, and transportation systems.

In July, Gov. Glenn Youngkin took two major steps on AI. First, he signed an executive order to streamline state regulations and guidance documents. Second, he launched the AI Career Launch Pad, a platform developed in partnership with Google to help Virginians pursue careers in artificial intelligence. The hub is housed on the Virginia Has Jobs website, originally created to help federal workers find employment opportunities in the commonwealth. 

According to the governor’s office, the AI site offers curated, no-cost and low-cost learning opportunities, ranging from beginner-friendly courses on AI fundamentals and practical workplace applications to boot camps and degree programs through Virginia’s community colleges and universities. Virginia Works – the state’s new workforce development agency – will also distribute disburse career certificate scholarships to expand access to AI training programs.

Earlier in the year, Virginia’s Department of Transportation announced plans to use AI to improve cost estimates and pavement management, citing a 68% nationwide increase in highway construction costs since 2020.