A Senate Bill Proposes US Government Control over AI Model Releases
Senate negotiators are drafting legislation that would empower the U.S. government to block the release of potentially unsafe AI models, according to Reuters. The bill aims to establish a duty of care for developers and grant the government the right to intervene in model releases.
Key Points:
- Duty of Care: The proposal requires AI developers to design their products with catastrophic risk prevention as a primary goal.
- Government Intervention: The government would have the power to block specific models deemed unsafe, with the option for companies to challenge these decisions in court.
- Scope: The bill targets advanced AI models held by U.S. companies like Google, Anthropic, and OpenAI.
- Negotiators: Three senators—John Thune (Senate majority leader), Ted Cruz (chair of the Senate Commerce Committee), and Amy Klobuchar (senior Democrat)—are leading the talks.
- Testing Disputes: The legislation sparks debate over who should conduct safety tests—companies themselves or federal agencies like national laboratories. Maria Cantwell, ranking Democrat on Commerce, opposes the current draft, favoring model submission to these agencies for assessment.
- State Law Override: A controversial clause in the bill would override state laws regarding AI model risks.
This development highlights the ongoing debate over regulating AI safety and underscores the complex challenges surrounding the governance of advanced artificial intelligence.