Bill Gates proposes 'human-reserved' jobs and a token tax
In his first major essay on AI in three years, the Microsoft co-founder stops hand-wringing and offers a concrete playbook for a labor market and set of institutions he says are nowhere near ready.
Bill Gates published a roughly 6,000-word essay arguing that governments are unprepared for the scale of the AI transition, and paired the warning with two concrete proposals. "Many jobs will disappear forever," he writes, calling on governments to designate "human-reserved" jobs — roles he likens to nature reserves, where automation is possible but society chooses not to deploy it because the cost is too great. His reference point is personal: caregivers who tended to his father before he died of Alzheimer's, a job where, Gates says, "no robot could or should have done it." He also floated a tax on AI tokens, a mechanism meant to slow adoption and cushion displaced workers — an idea economists have debated for years but one that now gets a stamp from one of the industry's architects.
The essay is as much about institutions as individuals. "The highest priority is a monumental task: creating a domestic and international framework for dealing with AI," Gates writes, comparing the work to the US government's post-9/11 reorganization and warning that AI touches national security, employment, taxation, energy, elections, health, law enforcement, and more. He argues for US–China cooperation and told reporters his staff are arranging a November meeting with Xi Jinping to discuss governing the risks. It is a markedly more policy-specific turn from Gates, who has cautioned about AI's dangers before but now attaches concrete, arguable proposals — the token tax especially is certain to draw pushback from labs that see compute taxes as a throttle on frontier research.
Beijing issues first penalty tied to an AI trade-secret leak
China's regulators handed down what is billed as the country's first administrative penalty in an artificial-intelligence trade-secret case, fining an individual 100,000 yuan for copying an AI software system and leaving to start a rival venture. Beijing's market supervision authority, which announced the ruling, said it acted under Article 26 of the Anti-Unfair Competition Law. Regulators also told the affected company to close the internal gaps the case exposed — tightening permission tiers, keeping logs, and auditing departing employees, the leak-vector the case was built around.
The case matters as a precedent: AI models, training pipelines, and proprietary algorithms have historically been hard to protect as trade secrets because they are intangible, technically intricate, and easily destroyed as evidence — precisely the traits the Beijing authority cited in its ruling. By treating a copied AI software system as a protected commercial secret and imposing a fine, Chinese regulators are signaling that AI assets get real legal cover even before formal AI-specific legislation settles in. Enforcers framing AI code as protected trade capital, and starting with the most concrete leak path — an employee walking out the door — gives companies a clearer line to lean on. Regulators turning to trade-secret law is notably lighter-touch than a criminal case, but it establishes the enforcement floor.
If entire job categories deserve "human-reserved" protection from AI, where would you draw that line? Tell us in the comments.
Sources: The Guardian · CNBC · Fortune · Beijing Municipal Market Supervision Administration · EO (经济观察网) · TMTpost (钛媒体)
