Bill Gates, the founder of Microsoft, warned that the rapid development of artificial intelligence could turn the technology into a broad source of disruption and inequality if it is not accompanied by appropriate oversight and regulatory rules. He expressed these views in an article published on his website, as well as in a recent interview with The New York Times, during which he said he had moved from early optimism about AI's potential to concern over the speed of its progress compared with the readiness of societies and infrastructure.
Gates's concern is based on the accelerating development of AI tools, including coding tools such as Anthropic's Claude Code, as well as conversational models that have begun to outperform humans in certain capabilities. In his view, the impact of these models is not limited to automating specific tasks, but could extend to both office and manual jobs, potentially making it more difficult for workers who lose their jobs to transition to other sectors.
From Job Losses to the Risks of Dangerous Use
Gates criticizes what he describes as the silence of some AI company leaders about the technology's potential risks. According to the material, these risks include the loss of large numbers of jobs, as well as the possibility of using AI to produce biological weapons. Gates believes that some companies may avoid discussing these issues publicly out of concern that doing so could affect investments or high valuations, amid the race to develop more advanced models.
These positions do not prove that such scenarios will occur, but they reveal the expanding scope of the debate, from improving productivity to distributing the effects of automation and defining the acceptable limits for using advanced systems. They also confront model developers with a question that goes beyond technical performance: How should the social and security consequences of the technologies they release be managed?
What Does Gates Propose in Practice?
Gates proposes imposing taxes on the basic computing units used to run AI models, which the material refers to as “tokens,” in addition to taxes on robots used in workplaces. According to his proposal, the aim is to make replacing employees with automated systems more costly, then direct the revenue toward retraining affected workers and preparing them for new jobs.
He also calls for creating areas “reserved for humans” in which AI systems would not perform certain tasks requiring deep human interaction. The examples he mentioned include mental-health specialists, caregivers, and doctors responsible for informing patients of sensitive medical diagnoses.
At the international level, Gates proposes creating a specialized oversight body to address cross-border risks, particularly the possibility of using AI to develop biological weapons, modeled on existing international bodies and agreements in areas such as nuclear energy and aviation.
Why Does This Debate Matter?
The practical value of Gates's proposal does not lie in its being a complete regulatory plan, but in linking three paths that are often discussed separately: the impact of automation on jobs, the responsibility of the companies developing it, and threats that do not stop at the borders of a single country. If these ideas become policy, they would affect model companies, users of AI tools, and workers whose jobs depend on cognitive or operational tasks that can be automated.
Nevertheless, the feasibility of the proposals remains disputed. Computer science professor Oren Etzioni told ABC News that imposing a tax on AI units could be like imposing a tax on keystrokes on a typewriter. He warned that imposing a tax in the United States could lead companies to use foreign models, giving competitors outside the country a competitive advantage. Some experts also believe that developing policies through existing government agencies may be more practical than creating new bureaucratic bodies.
Editorial reading: What has genuinely changed is the debate's shift from the question of what AI can do to the question of how its costs and risks should be distributed. The open questions concern how to define an “AI tax,” measure its impact, prevent usage from shifting to external parties, and identify the jobs that merit human protection. The material presents these positions and objections, but does not provide a detailed model for implementing the proposals or evidence that conclusively determines their outcomes.