Humanoid robot
An Anthropic safeguards researcher warns that superhuman AI could eventually eliminate human employment, potentially leaving people reliant on government welfare Magda Ehlers/Pexels

An Anthropic researcher has warned that artificial intelligence could eventually make human employment obsolete, leaving people dependent on government welfare as machines outperform workers across professions. Tao Lin, who works on AI safeguards at Anthropic, made the prediction in an interview with Palisade Research, a non-profit studying AI risks.

Lin described a future in which workers lose their current jobs, briefly move into different occupations and then discover there are no roles left for humans. 'Pretty much all humans on Earth will become permanently unemployable and will need to strictly rely on government welfare to sustain themselves. Human employment will go away,' he said.

The comments represent Lin's personal forecast, not an official prediction from Anthropic.

Superhuman AI Could Arrive Within Years

Lin estimated that broadly superhuman AI could emerge in approximately 1.3 years, potentially around the end of 2027. He described it as technology capable of matching or exceeding the best human experts in almost every field.

That could include areas such as programming, filmmaking and politics, potentially putting specialists under pressure.

However, Lin did not suggest mass unemployment would begin as soon as such systems appeared. There would be a gap between AI becoming capable of performing a job and businesses having the infrastructure to replace human workers.

Robots Remain a Major Obstacle

Physical work could take longer to automate because there are not yet enough capable robots to replace people on a massive scale.

Lin argued that companies would first need to improve their ability to manufacture robots before producing the billions potentially required across industries. Jobs involving physical tasks may remain available even if AI becomes vastly more capable in digital and intellectual fields.

Office workers and people in computer-based professions may face disruption earlier, while industries dependent on physical labour could take longer to change. Welfare payments and other forms of income support could become essential if people cannot earn a living through traditional work. The scale of such a shift would place enormous pressure on public finances and force governments to reconsider how economic gains are distributed.

Researcher Calls for Slower AI Development

Lin believes AI companies should dramatically reduce the speed of their research, proposing that development proceed at roughly a quarter of its current pace. He argued that slowing the race would give researchers more time to improve safety measures and assess risks.

His concerns extend beyond jobs. Lin discussed the possibility of advanced AI acting against human interests and estimated a 20 per cent chance that AI could eventually cause humanity's extinction. He also described a scenario in which people remain alive but lose meaningful control over a world run by machines.

Anthropic has separately acknowledged that AI could create serious economic disruption. In its economic policy framework, the company said advanced AI could reduce demand for human labour and proposed preparations ranging from workforce training and unemployment support to basic income or other ways of sharing the gains.

Lin's comments should therefore be understood as one researcher's assessment of a possible outcome, rather than a timetable or certainty.

For workers, the central question is whether governments, employers and communities can adapt quickly enough to ensure AI's benefits are widely shared. The forecast remains uncertain, and other researchers may disagree with Lin's timeline or assumptions about future labour markets.