Bill Gates is calling for AI labs to implement systems that can track what their models are doing instead of relying on a "kill switch" to prevent serious harm.
Speaking in an NBC interview aired on Sunday, the Microsoft co-founder said he wasn't against a system that would take models offline in the event of a disaster. He argued that a kill switch has a place but would not be enough in an era where "small groups with AI can do what only the biggest countries could do."
“I would never want to say I’m against a kill switch,” he added. “But when you’re trying to moderate the bad behavior, you need insight and records of what’s being done. So the kill switch alone would not prevent these tragedies.”
Malicious actors can leverage AI to attack power grids or break into bank accounts, Gates said, adding that the technology has also reached key thresholds in areas including biotechnology and cyberattacks.
Meanwhile, he called on governments to step up regulation, saying they should be involved in overseeing what AI systems do. He noted that those measures would not have to slow the industry’s development or the U.S. race with China.
The philanthropist also warned that reaching an international agreement on AI rules could prove harder than Cold War negotiations over nuclear weapons.
“This is more difficult than that was,” he said, citing the complexity of getting countries to agree on rules for a technology developed largely by private companies.
He pointed to international agreements on aviation safety, the ozone layer, and nuclear weapons as examples of areas where countries had put aside their differences to establish common rules, arguing that AI would require a similar effort.