A shutdown mechanism, built by the developer
Senator John Kennedy, a Louisiana Republican, is introducing a bill on Wednesday that would require developers of artificial intelligence systems to build a shutdown mechanism into them. He is bringing it under unanimous consent, the procedure that lets a single senator’s objection stop it dead.
The design choice that distinguishes it from other kill-switch proposals is who holds the switch. Kennedy has been explicit that his version would not put the federal government in direct control of shutting a system down. The Next Web reported that the calculation behind that is a veto count: a bill read as expanding federal power over AI companies would be unlikely to survive this president.
As introduced, the measure has no cosponsors, no committee date, no floor date and no companion bill in the House.
Three bills, no timetable
Kennedy’s is one of three AI measures now in front of the Senate. Senator Josh Hawley has a bipartisan bill covering national security, civil liberties and labour protections, already introduced. Senator Bernie Sanders is briefing on a bill to prohibit the development of superintelligent AI outright, which he said on Tuesday he will introduce next week with Representative Greg Casar.
None of the three has been scheduled for anything. Lawmakers from both parties have told the Associated Press there is no timeline for a vote on any of them. Senator Mark Warner, a Virginia Democrat, put the institutional problem plainly: “We never do anything until we have a holy heck moment.”
Separately, Senate Majority Leader John Thune is working with Senator Amy Klobuchar on legislation that would give Washington authority to block unsafe models — a different mechanism from a developer-held switch, and the only one of these efforts with leadership behind it.
The politics around it
The administration is not asking for any of this. President Trump spent the week calling the push to slow AI development a hoax and a conspiracy, and on Tuesday Nvidia’s Jensen Huang told a conference in San Francisco that no new AI laws are needed at all. House Speaker Mike Johnson has said he expects Trump to meet AI platform leaders at the White House in the coming days.
Against that, the case for statutory shutdown requirements has been made loudest by the labs. OpenAI confirmed on Monday that it has been in safety talks with Anthropic and Google for weeks, and has backed embedded auditors in the House FRONTIER Act.
What to watch
Unanimous consent is a fast route and a fragile one: it takes one objection. If Kennedy’s bill is blocked on Wednesday, the useful information is who objects and on what grounds, because that is the first recorded Senate position on a developer-held kill switch. If it passes, it becomes the first AI safety requirement to clear either chamber.
