AI Kill Switch Act would let Trump admin order shutdown of rogue AI systems - Ars Technica

Summary

The Artificial Intelligence Shutdown Act proposes granting the U.S. Department of Homeland Security authority to shut down AI systems that could cause catastrophic outcomes, and it includes penalties for non-compliance.

Why it matters

The law aims to mitigate the unpredictable or malicious behavior risks of advanced AI models that have emerged from recent incidents involving OpenAI and Anthropic. By establishing a legal framework for mandatory shutdowns, it seeks to protect national security and public safety while encouraging companies to integrate reliable termination mechanisms.

Highlights

• Empowers the Secretary of Homeland Security, in consultation with Commerce and Intelligence officials, to slow down or completely shut down AI systems.
• Imposes daily fines of up to $20 million on developers who fail to comply with shutdown orders.
• Targets companies that earn at least $500 million in AI revenue or use $100 million worth of computing power, reflecting a focus on large-scale and high-impact models.

Critical take

The bill grants the executive branch sweeping authority without providing clear technical criteria for defining 'rogue' behavior, which could stifle innovation and raise concerns about overreach and legal due process.