IL SB3444

Creates the Artificial Intelligence Safety Act, establishing liability protections and transparency requirements for developers of frontier AI models in Illinois

Defines "frontier model" as AI trained using more than 10^26 computational operations or with compute costs exceeding $100,000,000

Shields developers from liability for "critical harms" (death/serious injury to 100+ people or $1 billion+ in property damage) if they publish safety protocols and transparency reports on their websites and did not intentionally or recklessly cause the harm

Allows compliance through alternative pathways: agreeing to EU Artificial Intelligence Act requirements under Article 56, or entering a federal agency agreement for model access and evaluation

Automatically sunsets if federal government enacts overlapping frontier model regulations, and recognizes compliance with substantially similar state frameworks