Bill Gates Calls for ‘Human-Reserved’ Jobs and AI Taxes as Automation Risks Acceleration

That would mark a major change in how governments approach the technology.

Gates wants some jobs protected by law

One of Gates’s most striking proposals is what he calls “Human Reserved” areas.

The idea is simple: some jobs should remain human even when AI can technically do them.

Gates points to healthcare, childcare, education, psychological support, and judicial proceedings. In these areas, governments could restrict or ban the use of AI to replace people in roles that require trust, responsibility and human interaction.

It is a very different approach from the usual argument that companies should be free to adopt whatever technology makes them more productive.

Gates compares the idea with nature reserves. Some parts of the economy, in his view, could be deliberately protected from technological development because society would lose something important if machines took over completely.

That could mean accepting less efficiency in exchange for keeping people in jobs that are considered socially important.

The proposal also points to a growing concern about where the next wave of automation will hit. Previous technological revolutions often replaced repetitive work while creating new jobs elsewhere. AI is different because it can work with information, language and digital tasks.

That puts software developers, lawyers, financial workers, administrators and customer support staff in the firing line as well.

A tax on AI could change the calculation

Gates also wants governments to make automation more expensive.

His idea is a “token tax” - a small charge on the basic units of data processed by AI systems. He also supports taxes on robots and other forms of automated capital.

The purpose would be twofold. First, it would make companies think twice before replacing large numbers of workers with AI. Second, the money could help pay for retraining and social programmes for people who lose work.

The proposal goes to the heart of a problem that European governments are likely to face if AI becomes much more productive.

Today, a large part of public revenue is linked to human employment. Workers pay income taxes and social contributions, while employers contribute through payroll-related taxes. If machines gradually perform more of the work, governments could face a smaller traditional tax base even as the economy becomes more productive.

Gates's answer is to tax part of the infrastructure behind that productivity.

For companies, however, the effect would be immediate: AI would no longer be simply a tool that cuts labour costs. Some of those savings would be taxed.

That could also favour the biggest technology companies. Large firms are better placed to deal with complicated regulations and expensive compliance requirements than small AI startups.

Europe has not seen the jobs crisis yet

There is an important reason not to treat Gates’s warning as a description of what is already happening.

European data does not show mass AI-driven unemployment.

According to data from European institutions through mid-2026, AI is being used by 52 per cent of European workers, compared with 26 per cent in 2024. Companies using AI are reporting productivity improvements, but there has not yet been a broad fall in employment as a result.

For now, AI is mostly changing what people do during the working day rather than removing their jobs altogether.

That does not mean the labour market is standing still.

Demand for AI skills has increased sharply. Employer demand for “AI fluency” grew fivefold between late 2023 and late 2025, appearing in job advertisements covering about 5 per cent of total employment.

The immediate picture therefore looks much less dramatic than Gates’s warning.

The longer-term forecasts are another matter.

McKinsey estimates that 58 per cent of working hours in Europe could technically be automated by 2030. About 44 per cent could be affected by cognitive AI agents and another 14 per cent by physical robotics. The potential economic value is estimated at up to $1.9 trillion.

The same projections suggest that around 6.5 per cent of European workers - roughly 12 million to 14 million people - could eventually need to change occupations completely.

That is the gap between the present and the future that governments have to think about.

The workplace could get much smaller

Mark Zuckerberg has offered a very different vision.

The Meta chief executive expects small teams using AI agents to handle complicated work that currently requires much larger corporate organisations.

From a business perspective, that is an attractive prospect. A company could potentially produce more with fewer employees.

For workers, however, the consequences depend on what happens to the jobs that disappear.

The most vulnerable may not necessarily be people doing the simplest physical tasks. AI is increasingly aimed at work that was once considered difficult to automate because it required education, language skills or professional knowledge.

A junior programmer, legal associate, administrative worker or customer support employee may find that the entry-level position that once served as a route into a profession is no longer needed in the same numbers.

That creates a problem that productivity figures alone cannot capture.

An economy can become more efficient while becoming harder to enter for younger workers.

Gates is also treating AI as a security problem

Employment is only one part of Gates’s concern.

He argues that governments need stronger controls over AI systems that could create national security risks. Among the dangers he identifies are biological weapons, deepfake-driven fraud and autonomous systems that become difficult for humans to control.

He wants high-risk models to undergo mandatory safety reviews, particularly when they can be used for both civilian and military purposes.

Biological applications are especially sensitive. AI systems that can help design or work with chemical or biological pathogens raise risks that cannot be handled simply as another technology-sector safety issue.

Gates believes the US and China need to cooperate on these problems despite their wider strategic rivalry.

His team is arranging a summit in Beijing for November 2026 involving Chinese President Xi Jinping. The aim is to discuss global AI risks and possible forms of cooperation.

That is an ambitious idea. But Gates's argument is that the most dangerous AI capabilities will not respect national borders, so purely national regulation will have limits.

He has compared the scale of oversight he wants with the US government's major security reorganisation after September 11.

Whether governments can build anything comparable is another question.

Silicon Valley has a problem with incentives

Gates’s strongest criticism is aimed not at AI itself but at the business model surrounding it.

Technology companies are spending enormous amounts on data centres and computing infrastructure. They also have strong reasons to convince investors that AI adoption will continue to grow.

Gates argues that this creates a conflict of interest. Companies that have invested heavily in the next generation of AI have a financial reason to emphasise its benefits and play down the risks.

His claim that executives are privately more worried than they admit publicly is difficult to measure from the outside. But the underlying conflict is easy to understand.

The companies building the technology are also among those with the most to gain from rapid deployment.

That is why Gates’s call for government intervention is more significant than another warning from an AI critic.

He is arguing that the people developing the technology cannot be left alone to decide how quickly it should spread.

Europe will have to deal with the gap

For now, European governments have time.

AI has already reached a large part of the workforce, but the employment figures do not yet show the kind of disruption Gates fears. Companies are mostly using the technology to change tasks and increase productivity.

The problem is the speed at which that situation could change.

If AI agents begin taking over larger parts of professional work, governments may find themselves dealing with declining demand for human labour at the same time as they need more money for retraining and social protection.

That is where proposals such as a compute tax become politically important.

So does the idea of protecting certain jobs by law.

Neither policy is guaranteed to work. Both would interfere with the normal economic incentive to automate whenever automation is cheaper or more productive.

But that is precisely Gates’s point.