The Senate AI and Social Media Subcommittee met Thursday for the first of two days of hearings on measures related to artificial intelligence (AI), consumer protection and data centers. 

Sen. Bill Cunningham (D-Chicago) — chair of the committee — said the goal of the hearings is to listen to testimony on the bills proposed in the area, with the goal of finding the best path forward on the topic areas.  

During a panel on AI, Sen. Mary Edly-Allen (D-Grayslake) presented three measures that aim to protect children and set failsafes to mitigate risks to cybersecurity.  

Senate Bill 3261 — dubbed the Artificial Intelligence Public Safety and Child Protection Transparency Act — would require large developers to create, comply with and make publicly available a public safety and child protection plan.  

The measure would apply to chatbot developers that make at least $25 million in revenue annually and frontier developers that make at least $500 million in revenue annually. Frontier developers is a designation for those working with advanced general-purpose models.   

It would also require the Illinois attorney general to create a mechanism by which developers or members of the public can report a safety incident related to AI.   

A frontier developer would be required to report a critical safety incident within 15 days and report a critical safety incident that poses an imminent risk of death or serious physical injury within 24 hours. A large chatbot provider would have 15 days to report a child safety incident. 

A companion measure — Senate Bill 3262 — would restrict AI products from implementing certain features for minor users unless an adult has given permission.  

Edly-Allen’s third bill, Senate Bill 3312, would require frontier AI developers to adopt and publish a framework addressing catastrophic risk management, transparency and cybersecurity. It would also mandate developers to report critical safety incidents to the Illinois Emergency Management Agency and Office of Homeland Security.   

Senate Bill 3368 — sponsored by Sen. Sue Rezin (R-Morris) — aims to hold companies liable if an AI chatbot provides incorrect or dangerous information that results in physical or financial harm. The measure would also require companies to disclose when a user is speaking to an AI chatbot instead of a human representative.   

The measure also requires parental consent for a minor to use a chatbot. It further requires companies to use “technically feasible methods” to prevent chatbots from encouraging self-harm and would require chatbots to shut down and display resources for crisis support if a user expresses suicidal or self-harm intentions.   

Rezin said the ultimate goal is to protect kids who are using AI chatbots.  

“We're seeing an explosion in AI companions and AI friends without guardrails,” she said. “These programs can foster unhealthy dependencies or even encourage self-harm in children. There are several instances where families and children have used their AI companion and ultimately committed suicide and have settled out of court with the companies.”  

Senate Bill 3384 — sponsored by Sen. Laura Ellman (D-Naperville) — would also require companies to establish protocols for AI chatbots to prevent them from encouraging suicidal ideation and to refer people to the proper sources if they do express a desire to self-harm.  

Sen. Rachel Ventura (D-Joliet) presented Senate Bill 2502. The measure would define AI as a product and establish standards for transparency, accountability and oversight, including a framework to evaluate risks and require safeguards.  

Ventura said the measure is needed because of the speed at which AI is growing. She said AI has started to have negative impacts on everyday life.  

“That speed is outpacing our ability to understand and manage the risk, let alone legislate after the fact,” she said. “Without guardrails, we are effectively beta testing powerful systems on the entire public. We are seeing real world consequences, [such as] routine automation amplifying bias in screening of qualified candidates from employment to housing and lending, resulting in discriminatory outcomes.”  

Bethany Abbate, director of AI policy for the Software and Information Industry Association, said adopting a product framework for AI would be harmful, as the technology is constantly evolving and responding to the prompts people put in.  

She further said that making the ideas that AI producessubject to liability “would undoubtedly violate the First Amendment by restricting the public's right to access and receive information.” 

Cunningham also presented his bill, Senate Bill 3444.   

The measure is an initiative of OpenAI, the AI research and development company behind ChatGPT, and would require companies to either adhere to safety and security requirements adopted by the European Union or enter into an agreement with a federal agency to establish safety and security requirements. The company would have to publish these requirements and a transparency report to its website.  

The measure also states that if a frontier AI company has done the above, they will not be held liable for “critical harms caused by the frontier model if the developer did not intentionally or recklessly cause the critical harms.”   

Jarrett Catlin, state AI policy advisor for TechNet, said states like Illinois passing their own laws creates a patchwork environment, which can be difficult for smaller companies to keep up with. He said Illinois also risks becoming a “compliance outlier,” which could result in companies ceasing to operate in the state entirely.  

Scott Wisor, policy director for the Secure AI Project, said that as AI continues to develop, regulations are essential to ensure safety.  

“At least seven families have filed lawsuits against AI companies after they lost children to suicide following engagement with AI chatbots,” he said. “At least three of the leading frontier AI companies have disclosed in their own safety evaluations that their models may provide meaningful uplift to novices seeking to build biological weapons. Multiple cyber-attacks have been reported in the last year to have been enabled by advanced AI systems.”   

He said polling showed that most Illinoisans support regulations for AI, especially as they relate to safety for minors.  

The committee will meet on Friday at noon.