A Massachusetts Democrat is calling on the United States to pass legislation that would move artificial intelligence away from nuclear power.
On Thursday, Senator Edward Markey said: “78 years ago this weekend, Robert Oppenheimer witnessed the world’s first explosion of nuclear weapons. In 2023, we face a new type of nuclear threat: the weaponization of increasingly powerful artificial intelligence systems.
“We need to pass legislation to get AI off the nuclear button before it’s too late,” he said.
Markey’s office said he filed more than a dozen amendments to the National Defense Authorization Act, including language that would ban the use of AI in nuclear launch decisions. of the US Army.
LEGISLATIVE EDGE BY AI-LAUNCHED NUKES, DEMAND ‘HUMAN CONTROL’ IN DEFENSE POLICY BILL
Additionally, the office said the amendments would advance nuclear disarmament and nonproliferation and save billions of federal dollars by shifting wasteful spending from nuclear weapons development to vaccine research.
One amendment is based on the Block Nuclear Launch by Autonomous Artificial Intelligence Act, which would ban AI from making nuclear launch decisions.
Markey and bipartisan Representatives Ted Lieu, Don Beyer and Ken Buck introduced the bill in April.
AI BANNED FROM OPERATING NUCLEAR MISSILE SYSTEMS UNDER BIPartisan BILL
“We need to keep humans informed about life-and-death decisions to use lethal force, especially for our most dangerous weapons,” Markey said then.
“While US military use of AI may be appropriate to enhance national security, the use of AI to deploy nuclear weapons without a human chain of command and control is reckless, dangerous, and should be prohibited” , said Buck, R-Colo.
CLICK HERE TO GET THE FOX NEWS APP
Their bill would codify existing Pentagon policy that requires a human being to be “informed” of any decision regarding the use of nuclear weapons.
In its fiscal year 2024 budget, the Pentagon is asking for $1.8 billion just for research and development of AI capabilities.
Fox News’ Elizabeth Elkind contributed to this report.