Gates urges US AI monitoring law beyond a shutdown switch
In a September 27 NBC interview, Bill Gates called for US law requiring AI safeguards and monitoring, arguing that company self-regulation and an emergency shutdown control are insufficient. He focused on misuse by people with harmful intent, including cyberattacks. This was a policy call, not an enacted rule or a finding from a safety test.
Artificial Intelligence··Morning
Gates calls for binding US safety rules
Microsoft co-founder Bill Gates used a September 27 NBC Meet the Press interview with Kristen Welker to call for US law on AI safeguards. He said legislators and law enforcement should help define the monitoring that companies would be required to carry out. Leaving the issue to voluntary company choices was insufficient in his view. Gates argued that the extra work for developers would be modest and need not greatly slow the industry. His proposal concerns how AI systems are overseen; the interview itself did not enact a rule or announce a completed safety programme.[1], [2]
A shutdown control does not reveal harmful use
Gates did not reject an emergency control that can turn off a model. He argued that such a switch alone would not show what a system had done or allow harmful behavior to be monitored. The danger he emphasized was not a model already taking over computers and resisting shutdown. He pointed instead to malicious people using available tools for cyberattacks, interference with bank accounts or attacks on power grids. Those examples were risks he described, not incidents established by the interview. Monitoring and records, in his argument, concern actions taken before a shutdown.[1], [2]
His extreme warning is a hypothetical outcome
Gates warned that AI in hostile hands could let small groups inflict harm on a scale once associated with large countries. He raised the possibility of events leading to a billion deaths. That figure was a warning about a possible catastrophe, neither a death toll nor a calculated probability. Gates also acknowledged President Donald Trump’s concern that regulation could slow US AI development against China, but said he hoped further discussion could find agreement. He mentioned aviation safety and nuclear arms arrangements as examples of international cooperation. No safety test, negotiated agreement or legislative timetable was presented in the interview.[1]
