On April 10, OpenAI announced its support for Illinois bill SB 3444. The bill stipulates that if an advanced AI model causes "catastrophic harm," the developer shall be exempt from liability provided they did not cause the harm through intent or gross negligence and had published safety, security, and transparency reports on their website.
The bill defines "advanced models" as AI models trained with a computational cost exceeding $100 million, a category that could include OpenAI, Google, xAI, Anthropic, and Meta. Examples of "catastrophic harm" cited include casualties of 100 people or more, property damage exceeding $1 billion, or the production of chemical, biological, radiological, or nuclear weapons using AI.
Jamie Radice, a spokesperson for OpenAI, stated that the bill "focuses on reducing the risk of catastrophic harm from cutting-edge AI systems while enabling this technology to reach the people and businesses of Illinois," and added that it supports a transition toward consistent national standards to avoid a patchwork of state-by-state rules.
According to AI policy experts, this bill represents a more aggressive measure than those OpenAI has supported in the past and indicates a turning point in its legislative strategy.
Source: OpenAI backs Illinois bill that would limit when AI labs can be held liable（HN 447pt・324コメント）（HN Search (backfill）)