May 24, 2024
Tort Law and Frontier AI Governance
The development and deployment of highly capable, general-purpose frontier AI systems—such as GPT-4, Gemini, Llama 3, Claude 3, and beyond—will likely produce major societal benefits across many fields. As these systems grow more powerful, however, they are also likely to pose serious risks to public welfare, individual rights, and national security. Fortunately, frontier AI companies can take precautionary measures to mitigate these risks, such as conducting evaluations for dangerous capabilities and installing safeguards against misuse. Several companies have started to employ such measures, and industry best practices for safety are emerging.
Frontier AI developers can take precautions to reduce the risks that their most advanced systems will increasingly pose.
It would be unwise, however, to rely entirely on industry and corporate self-regulation to promote the safety and security of frontier AI systems. Some frontier AI companies might employ insufficiently rigorous precautions, or refrain from taking significant safety measures altogether. Other companies might fail to invest the time and resources necessary to keep their safety practices up to date with the rapid pace at which AI capabilities are advancing. Given competitive pressures, moreover, the irresponsible practices of one frontier AI company might have a contagion effect, weakening other companies’ incentives to proceed responsibly as well.
Read the entire article from Lawfare.
More from CNAS
-
AI and the Evolution of Biological National Security Risks
Reports
Executive Summary Not long after COVID-19 gave the world a glimpse of the catastrophic potential of biological events, experts began warning that rapid advancements in artific...
By Bill Drexel & Caleb Withers
-
The 1960s Novella That Got AI (Mostly) Right
Commentary
A secret military project. A vast artificial mind. Questions of consciousness. These form the premise of Dino Buzzati’s The Singularity, originally published in 1960 at the da...
By Paul Scharre
-
Response to NTIA Request for Comment: “Dual Use Foundation Artificial Intelligence Models with Widely Available Model Weights”
Commentary
In February 2024, the Department of Commerce’s National Telecommunications and Information Administration (NTIA) issued a Request for Comment (RFC) on the implications of “ope...
By Caleb Withers
-
Future-Proofing Frontier AI Regulation
Reports
Executive Summary Policymakers should prepare for a world of significantly more powerful AI systems over the next decade. These developments could occur without fundamental b...
By Paul Scharre
