OpenAI and Anthropic Split Over Illinois AI Liability Bill

OpenAI and Anthropic are on opposite sides of a proposed Illinois law that could shape how AI companies are held accountable for severe damages caused by their models, reports Inc.. The bill, known as SB3444 or the Artificial Intelligence Safety Act, was introduced by state senator Bill Cunningham on February 4.

The measure would protect developers of frontier AI models from liability for “critical harms” if they can show the harm was unintentional and that they follow safety, security, and transparency protocols. “Critical harm” is defined as incidents resulting in at least 100 deaths or injuries, or $1 billion in property-related damages. Companies complying with the European Union’s AI Act would be automatically considered in compliance under the Illinois proposal.

The bill’s structure resembles the federal Protection of Lawful Commerce in Arms Act, which shields gun manufacturers from liability for unlawful misuse of their products. According to the report, OpenAI supports the measure, while Anthropic has lobbied for changes or for the bill to be withdrawn. OpenAI’s global affairs representative Caitlin Niedermeyer signed a witness slip in favor of the legislation and testified before an Illinois Senate subcommittee earlier this month.