OpenAI Supports Illinois Bill to Limit AI Liability

OpenAI is backing a controversial Illinois bill, SB 3444, which aims to protect AI labs from liability in cases where their models cause significant harm, such as mass casualties or extensive property damage. This marks a shift in OpenAI's legislative strategy, as it previously opposed similar bills. The bill would exempt AI developers from liability if they did not intentionally or recklessly cause harm and have published safety and transparency reports. The bill's definition of 'frontier models' includes those trained with over $100 million in computational costs, targeting major AI labs like OpenAI and Google.

OpenAI argues that this approach focuses on reducing risks from advanced AI systems while promoting consistent national standards, avoiding a patchwork of state regulations. The bill lists potential critical harms, such as AI creating weapons or committing criminal acts autonomously. Despite the bill's state-level focus, OpenAI's Caitlin Niedermeyer advocates for a federal framework to harmonize AI regulations, aligning with Silicon Valley's view that legislation should not hinder the US's global AI leadership.

However, the bill faces opposition in Illinois, known for strict tech regulations. Scott Wisor of the Secure AI project notes that public opinion strongly opposes exempting AI companies from liability. Illinois has previously increased AI liability, such as limiting AI in mental health services and regulating biometric data. The broader question of AI liability remains unresolved, with federal legislation still elusive despite efforts from the Trump administration and some state-level initiatives.

As AI technology advances, the debate over liability and safety continues, with states like California and New York requiring AI developers to submit safety reports. The potential for AI to cause catastrophic events remains an open legal question, highlighting the need for clear regulations.