Anthropic Researcher Warns AI Could Make Nearly All Humans Unemployable, Predicts Superhuman AI by Late 2027

news
Anthropic Researcher Warns AI Could Make Nearly All Humans Unemployable, Predicts Superhuman AI by Late 2027

An Anthropic researcher has warned that artificial intelligence could eventually replace almost every form of human employment, potentially leaving governments responsible for financially supporting much of the population.

Tao Lin, who works on safety measures for Anthropic's Claude AI models, believes broadly superhuman artificial intelligence could emerge by the end of 2027. Such systems would potentially match or exceed the capabilities of leading human experts across nearly every intellectual field.

However, Lin does not expect widespread unemployment to occur immediately after these systems become available. He believes the transition would happen gradually as AI replaces existing occupations and eventually becomes capable of performing tasks that currently require human workers.

His predictions, made during an interview published in mid-November 2026, represent his personal assessment rather than an official forecast from Anthropic.

Anthropic Researcher Predicts Superhuman AI Could Arrive Within 1.3 Years

During an interview with Palisade Research, Lin estimated that broadly superhuman AI could arrive approximately 1.3 years from the time of his forecast.

This would place its potential emergence around the end of 2027.

He described a future in which AI systems could outperform even highly skilled professionals in fields such as programming, filmmaking, and other demanding intellectual activities.

If such capabilities become available, companies could potentially automate a much larger share of their operations.

Lin believes this progression could eventually make human workers economically unnecessary across most occupations.

He suggested that people displaced from their existing jobs might initially find employment in other industries. However, those alternative positions could also become vulnerable as AI capabilities continue improving.

In his view, the process could eventually leave nearly all humans permanently unable to find economically competitive employment.

Such an outcome could require governments to develop more extensive financial support systems for people who can no longer earn wages through conventional employment.

However, Lin did not provide a specific date for when this level of unemployment might occur.

Robotics Could Delay the Replacement of Human Workers

One important limitation identified by Lin concerns the physical capabilities of automation.

AI systems may eventually handle complex intellectual tasks, but replacing people in physically demanding occupations would require significant advances in robotics and supporting infrastructure.

Factories, warehouses, construction sites, and other physical workplaces would need machines capable of performing tasks reliably and economically.

Even if advanced AI models become capable of planning and controlling these activities, deploying robots throughout the economy would take time.

Companies would also need to invest in equipment, infrastructure, maintenance, and changes to existing production systems.

This creates a potential delay between the emergence of superhuman AI and widespread automation of physical labor.

The distinction is important because advances in AI software do not automatically make every occupation replaceable.

Anthropic's Research Has Not Found Widespread AI Unemployment So Far

Lin's predictions differ considerably from the employment effects currently identified in Anthropic's own research.

In March 2026, the company published findings examining how AI exposure relates to employment trends.

The research found no systematic increase in unemployment among workers in occupations with substantial observed AI exposure since late 2022.

However, researchers identified tentative evidence of slower hiring among younger workers in certain occupations where AI tools are increasingly capable of performing relevant tasks.

This suggests AI could be affecting some employment opportunities without yet producing widespread job losses.

The research also identified a substantial difference between tasks AI can theoretically perform and those for which companies are actually using the technology.

For example, approximately 68% of observed Claude usage involved tasks considered theoretically achievable through AI alone.

That figure does not mean 68% of jobs have been automated or that an equivalent proportion of workers has become unnecessary.

Many occupations involve several responsibilities, and automating individual tasks does not necessarily eliminate an entire position.

Anthropic Announces $200 Million Research Fund to Study AI's Economic Impact

Anthropic has also acknowledged the possibility that AI could create longer lasting economic disruption.

In July 2026, the company published the research agenda for a $200 million research fund focused on examining the economic consequences of artificial intelligence.

The initiative includes research into unemployment insurance reforms, longer term income support, and policies intended to assist workers displaced by automation.

Such measures could become increasingly relevant if AI adoption significantly changes hiring patterns and labor demand.

However, the appropriate policy response would depend on how quickly AI capabilities improve, which industries adopt the technology, and whether new employment opportunities emerge.

The available evidence does not establish that universal unemployment is inevitable.

Researcher Raises Concerns About Losing Control of Advanced AI

Beyond employment, Lin also expressed concern about the safety of increasingly autonomous artificial intelligence systems.

He assigned an approximately 20% subjective probability to a scenario in which advanced AI eventually removes humanity from the equation, though he said he does not consider direct human extinction particularly likely.

This estimate reflects his personal assessment of the risks rather than a scientifically established probability.

Lin believes current AI development is progressing too quickly for safety measures and government policies to keep pace.

He has advocated reducing the speed of frontier AI research to approximately one quarter of its present rate.

According to his argument, slowing development would provide researchers and policymakers with more time to understand emerging risks and introduce appropriate safeguards.

He also acknowledged that people have reasons to question AI developers about technologies that could substantially change their employment prospects and everyday lives without their direct consent.

Superhuman AI and Universal Unemployment Remain Uncertain Predictions

Lin's comments highlight the concerns some researchers have about the long term effects of increasingly capable AI systems.

However, several major uncertainties remain.

Researchers have not established when broadly superhuman AI will become available, whether it will outperform humans across every occupation, or how quickly businesses could adopt such technology.

Economic factors, infrastructure limitations, regulatory decisions, and the continued demand for human involvement could all influence employment outcomes.

Anthropic's existing research indicates that widespread AI related unemployment has not yet appeared in the available occupational data, even as AI tools become increasingly capable.

The difference between Lin's long term prediction and current employment evidence is therefore significant.

His forecast raises questions about how governments and employers might respond to advanced automation, but it does not demonstrate that human employment will disappear or that superhuman AI will arrive by the end of 2027.

Discover: News

Discussion (0)

Be the first to comment.