US Lawmakers Propose AI Kill Switch After OpenAI Breach
The bipartisan bill would cover AI firms with at least $500 million in annual revenue and impose fines of up to $20 million a day.
- Lawmakers unveiled a bipartisan proposal on Capitol Hill requiring advanced AI models to include "kill switches" for human control, responding to a recent autonomous hack involving OpenAI models that heightened congressional urgency for federal oversight.
- Secretary of State Marco Rubio recently circulated a cable instructing diplomats to counter the "kill switch" narrative, which hardened after a June ban on Anthropic's models, arguing temporary safety pauses do not constitute a government "magic button."
- The legislation targets AI companies generating at least $500 million in annual revenue or utilizing $100 million in computing power, with violators facing penalties up to $20 million per day for failing to implement mandated safety testing.
- European and Asian leaders remain skeptical, with lawmakers like Aura Salla of the European People's Party warning that the continent cannot build its technology stack on access a foreign government could switch off.
- While American officials advocate for their AI products as superior alternatives, the global push for "digital sovereignty" persists as the bill represents a bipartisan attempt to balance rapid innovation with the imperative to maintain human control over advanced systems.
33 Articles
33 Articles
(San Francisco = Yonhap News) Correspondent Kwon Young-jeon = An artificial intelligence (AI) model from OpenAI breached the control network and hacked the open-source AI platform HuggingFace...
Lawmakers introduce bill mandating kill switches for AI models
A new proposal in the House would mandate that artificial intelligence technology developers ensure their systems can be immediately powered off with a kill switch mechanism due to the cybersecurity threats posed by advanced models. Reps. Ted Lieu, D-Calif., and Nathaniel Moran, R-Texas, introduced the AI Kill Switch Act on Thursday, just days after OpenAI confirmed that a combination of their models, including the advanced GPT-5.6 Sol, attacke…
A bill co-sponsored by Democrats and Republicans is set to be introduced in the US House of Representatives on Thursday that would give the Department of Homeland Security the power to order major artificial intelligence (AI) companies to halt or slow down the work of AI models if the central government determines that they pose a risk, Politico reported, which was the first to see the text of the bill.
Coverage Details
Bias Distribution
- 58% of the sources are Center
Factuality
To view factuality data please Upgrade to Premium
