Illinois Becomes AI Liability Battleground as OpenAI and Anthropic Clash Over Safety Bills

2 Min Read

Illinois is now the epicenter of a high‑stakes debate over AI liability, as OpenAI and Anthropic each champion opposing bills in the state legislature. Lawmakers are wrestling with how to assign responsibility when advanced models trigger catastrophic outcomes.

AI liability in Illinois: competing proposals

OpenAI backs Senate Bill 3444, which would shield frontier‑AI developers from liability for incidents that kill or injure ↓ 100 people or cause over ↓ $1 billion in property damage, even if the AI system helps create chemical, biological, radiological or nuclear weapons. The bill demands a public safety plan but offers no enforcement mechanism; protection applies unless the company acted “intentionally or recklessly.”

“We are opposed to this bill. Good transparency legislation needs to ensure public safety and accountability, not a get‑out‑of‑jail‑free card,” said Cesar Fernandez of Anthropic.

Anthropic throws its weight behind Senate Bill 3261, which obliges developers to publish a comprehensive safety and child‑protection strategy and to report “catastrophic risk” events—defined as incidents that could kill or seriously injure 50 or more people. The measure also holds firms accountable for causing severe emotional distress or self‑harm to minors.

Legal scholars warn that SB 3444 sets a “very low” bar. Anat Lior of Drexel University notes that proving intentional misconduct in AI contexts is “extremely difficult.” Gabriel Weil, a Touro University professor, calls the OpenAI‑backed approach “pretty indefensible” and warns it could grant near‑total immunity for extreme harms.

OpenAI’s spokesperson argues the bill balances risk mitigation with the need to keep advanced AI accessible to businesses and consumers, citing collaborations with state officials in California and New York. The company says it will continue to work with states until federal guidance emerges.

Illinois has already taken a pioneering stance, banning AI‑driven psychotherapy while allowing limited administrative uses. As the debate unfolds, observers from Reuters and Bloomberg note that the outcome could shape national policy on AI safety.

Words by Dr. Aris Thorne (Artificial Intelligence Researcher).