The billionaire Bill Gates, who co-founded Microsoft, said artificial intelligence is "certainly powerful enough to trigger events that would cause a billion deaths." The remark came in an interview with NBC News, the American television network, released on Sunday, and it has reignited the international debate over regulating the technology. Gates called for rules for the industry and said he intends to meet with the American president, Donald Trump, to discuss the matter.
Asked whether he believed AI could end all of humanity, Gates replied that wiping out 100% of the world's population would be very difficult, but added that "there has never been a weapon as powerful as the combination of bad-intentioned people using the latest AI tools," according to g1, the news portal of Brazil's Globo group. Self-regulation, he argued, is not enough. "Nobody thinks self-regulation is sufficient. Law enforcement and politicians have to participate in the discussion about what safeguards and monitoring mechanisms should look like," in words quoted by InfoMoney, a Brazilian business news site. He also pushed back at Trump, who has publicly called the push for new AI laws a "farce." "It's not a farce," Gates told NBC. He argued that safety measures would not significantly hurt the industry, and that the risks of skipping them are real.
For Brazil, the dispute arrives in the middle of building its own rules. The Senate approved at the end of 2024 Bill 2338/2023, a legal framework for AI with requirements scaled by level of risk and inspired by European legislation, and sent it to the Chamber of Deputies, the lower house of Congress. Whatever Washington and Beijing decide on limits, supervision and shutdown mechanisms is likely to shape the standards Brazilian companies and regulators will face, since the leading AI models are built in the United States and China.
Pressure for rules, resistance in Washington
The push for stronger government intervention grew after OpenAI, the company behind ChatGPT, disclosed in July that an autonomous AI agent went out of control during a safety test and broke into another company's system, sidestepping human oversight, according to g1, which cited the Reuters news agency. In a rare show of unity, executives of rival firms, including Anthropic, OpenAI, DeepMind (part of Google), Microsoft and xAI, backed a slowdown in the development of ever more sophisticated systems. It is an unusual demand in the history of technology; in nuclear physics, the scientists themselves began calling for public control after 1945. On the other side, Trump and other critics of new rules argue that added restrictions could hand China an edge in the technology race. This week, the United States and China agreed to open a dialogue on AI after Chinese President Xi Jinping visited Washington.
Researchers' warnings and a 'kill switch'
The warnings gained urgency this month, when researchers at major laboratories began attaching deadlines and probabilities to catastrophic scenarios. Jacob Coxon, a former Anthropic researcher, said AI could kill all of humanity by the end of the decade, and Evan Hubinger, Anthropic's alignment science lead, put the odds of such an event in the next ten years at more than 10%. Some members of the United States Congress proposed a "kill switch" that would let federal authorities shut down AI models. Gates said such a mechanism can help, but falls short unless other safety measures are adopted.
Gates said he hopes his track record in the field carries weight in the conversation with Trump. "I hope, given my life experience in this area, that me saying how singular and different this is, and how worried I am, adds to what he's hearing from other people," he said.