A provocative proposal
In their October 1 essay, Rauterberg and Sanga argued that the traditional protection of limited liability, which shields shareholders from losing more than their investment, should not fully apply to AI companies whose products could cause serious harm. Their argument rests on the idea that the features that make AI valuable to customers are also what make it dangerous.
“Customers demand the capabilities that make the product so dangerous.” — Gabriel Rauterberg and Sarath Sanga, in their Wall Street Journal essay
The authors contend that if AI companies can externalise the risks of their products while shareholders enjoy the profits, there may be too little incentive to invest in safety. They also discussed government measures, including export controls, as ways to manage the risks of powerful AI systems. The proposal drew sharp responses from readers who saw it as unfair or impractical.
AI as a weapon
Nolen said the authors' observation immediately reminded him of weapons. When a weapon is used to harm someone, he noted, the manufacturer is generally not held liable; the user is.














