California Democrat Ted Lieu and Texas Republican Nathaniel Moran introduced legislation Thursday that would require firms developing the most powerful AI models to also create the technical ability to “throttle, suspend, or shut them down.” The U.S. government would be able to order companies to hit the kill switch. (The Logic)

Talking point: The lawmakers said they’re worried about the dangers AI could pose, citing the cyber capabilities of the technology that frontier labs have recently developed; OpenAI’s systems recently broke containment and hacked Hugging Face, while Anthropic initially caged its Mythos model for being too good at exploiting software flaws. The Trump administration has already delayed and restricted access to AI tools over security concerns. The bill’s sponsors argue the U.S. government needs proper powers to tackle worrying incidents. But its prospects are unclear—the Trump administration has already been trying to play down international talk that it holds any kind of AI kill switch. And Silicon Valley has argued any slowdown imposed on its AI advances would just help Chinese developers, who wouldn’t stop.