Bill Gates Warns an AI 'Kill Switch' Won't Be Enough After Raising Billion-Death Risk

Bill Gates has warned that artificial intelligence could become powerful enough to help trigger events causing a billion deaths, while arguing that an emergency 'kill switch' would not by itself prevent catastrophic misuse.

The Microsoft co-founder said governments need mandatory safeguards, monitoring and records showing how advanced AI systems are being used.

In an interview with NBC's 'Meet the Press' that aired on Sunday, 27 September, Gates said AI was 'certainly powerful enough' to drive events resulting in a billion deaths.

Gates did not say that a billion people would die, but warned that AI could enable events resulting in that number of deaths if increasingly capable systems fall into the hands of people intent on causing mass harm.

Gates's comments sharpen a debate over whether governments can rely on technology companies to regulate themselves as AI becomes more capable.

He said self-regulation was insufficient and argued that politicians and law-enforcement authorities should help determine what safeguards and monitoring companies must implement.

Why Gates Says an AI 'Kill Switch' Is Not Enough

A kill switch would, in principle, give operators or authorities a mechanism to suspend an AI service when it presents a serious danger. Gates argued that waiting until an emergency develops would not address many forms of misuse.

'It's not enough to have a kill switch,' Gates told NBC. He said preventing dangerous behaviour requires 'insight and records of what's being done', meaning powerful AI systems would need safeguards and monitoring alongside an emergency shutdown mechanism.

Gates also stressed that current AI systems have not reached a point where they autonomously take control of computers in a way that makes them impossible to shut down. His concern is broader: people could deliberately use sophisticated AI tools to cause damage before authorities had reason to activate an emergency cutoff.

Gates Points to Cyberattacks and Bioterrorism

The billion-death figure needs to be understood in that context. Gates did not identify a particular AI system capable of killing a billion people today, nor did he attach a probability to such an outcome.

Instead, he has repeatedly focused on the possibility that AI could lower the technical barriers to destructive acts. In an August essay, Gates warned that AI could help criminals conduct cyberattacks and potentially make it easier to design a deadly new disease.

He identified hospitals, financial institutions, water systems and power grids among infrastructure that could become targets.

That concern also appeared in his January writing, when Gates said a non-government group using open-source AI to design a bioterrorism weapon could pose a greater threat than a naturally occurring pandemic. At the same time, he has stressed AI's potential benefits in healthcare, education and other fields.

AI Safety Debate Moves Beyond the Emergency Button

Gates's comments place the focus on safeguards and monitoring before an emergency develops, rather than solely on shutting down a dangerous AI system.

His proposed approach centres on monitoring sophisticated models, keeping records of their activity and establishing safeguards through government-backed rules rather than voluntary commitments alone.

Gates said such requirements would impose some additional burden on AI companies, but need not dramatically slow development.

The debate is also international, with Gates arguing that safeguards need not put the United States at a competitive disadvantage and that other countries, including China, have an interest in preventing AI-enabled attacks.

The central point of his warning is not that a billion-death AI disaster is inevitable. Instead, he argued that increasingly capable AI could give people with malicious intent more powerful tools, while a kill switch alone would not provide the safeguards and monitoring he believes are necessary.