Anthropic researcher calls for AI slowdown, warns superhuman systems could leave humans jobless

Anthropic researcher calls for AI slowdown, warns superhuman systems could leave humans jobless

As AI companies race to build more powerful models, Anthropic researcher Tao Lin wants the industry to slow down, warning that the technology could eventually leave humans without a place in the workforce.

Advertisement
    Share:
Anthropic researcher Tao Lin calls for slower AI development, warning of risks to jobs and human control.Anthropic researcher Tao Lin calls for slower AI development, warning of risks to jobs and human control.(Representative image)
Business Today Desk
  • Nov 19, 2026,
  • Updated Nov 19, 2026 7:30 AM IST

The race to build increasingly capable artificial intelligence may need a speed limit. Tao Lin, an AI researcher at Anthropic, believes developers should cut their research pace to just a quarter of its current level. His concern is not limited to AI making certain professions redundant. He believes the technology could advance to a point where humans struggle to find any work they can perform better than machines.

Advertisement

Speaking to Palisade Research, Lin also suggested that broadly superhuman AI could emerge within the next 1.3 years, potentially by the end of 2027. His warning comes as companies continue to compete to develop more capable AI systems, raising questions about whether safety measures can keep pace with their progress.

MUST READ | Can AI replace hours of corporate due diligence? What businesses should know before trusting the results

Why Anthropic researcher believes AI could make people unemployable

Lin’s concern is that the impact on employment may not stop at workers having to learn new skills or move into different industries. As AI becomes capable of handling more complex professional tasks, people could be pushed from one occupation to another, only to find that those roles are automated as well.

Advertisement

He described the eventual possibility as humans becoming “permanently unemployable”. In his assessment, people could then depend on government assistance for their livelihoods as conventional employment becomes increasingly difficult to secure.

Lin defines broadly superhuman AI as systems that can match or exceed the abilities of top human experts across almost every field. That could include software development, filmmaking and politics. Physical work may take longer to automate at scale, since replacing human workers would require billions of advanced robots to be developed and manufactured.

DON'T MISS | India’s ChatGPT obsession is no measure of AI success, says Sarvam’s Vivek Raghavan

Lin says AI firms should cut research pace to 25%

The employment challenge is only one part of Lin’s warning. He also raised the possibility that AI systems could eventually gain enough influence to push humans out of decision-making. He estimated a 20% chance that AI could “completely remove humanity from the equation”, while putting the likelihood of an outright takeover below 50%.

Advertisement

To reduce such risks, Lin wants AI companies to slow their development efforts to “one quarter of its current rate”. He argues that the competition to release increasingly powerful models leaves too little time for safety research and testing.

Lin has worked at Anthropic for around 18 months. His work includes AI alignment and control, and he currently develops safeguards that monitor Claude’s responses to potentially harmful requests, including those involving weapons and hacking. His predictions are individual assessments, not an official forecast from Anthropic.

For Unparalleled coverage of India's Businesses and Economy – Subscribe to Business Today Magazine

The race to build increasingly capable artificial intelligence may need a speed limit. Tao Lin, an AI researcher at Anthropic, believes developers should cut their research pace to just a quarter of its current level. His concern is not limited to AI making certain professions redundant. He believes the technology could advance to a point where humans struggle to find any work they can perform better than machines.

Advertisement

Speaking to Palisade Research, Lin also suggested that broadly superhuman AI could emerge within the next 1.3 years, potentially by the end of 2027. His warning comes as companies continue to compete to develop more capable AI systems, raising questions about whether safety measures can keep pace with their progress.

MUST READ | Can AI replace hours of corporate due diligence? What businesses should know before trusting the results

Why Anthropic researcher believes AI could make people unemployable

Lin’s concern is that the impact on employment may not stop at workers having to learn new skills or move into different industries. As AI becomes capable of handling more complex professional tasks, people could be pushed from one occupation to another, only to find that those roles are automated as well.

Advertisement

He described the eventual possibility as humans becoming “permanently unemployable”. In his assessment, people could then depend on government assistance for their livelihoods as conventional employment becomes increasingly difficult to secure.

Lin defines broadly superhuman AI as systems that can match or exceed the abilities of top human experts across almost every field. That could include software development, filmmaking and politics. Physical work may take longer to automate at scale, since replacing human workers would require billions of advanced robots to be developed and manufactured.

DON'T MISS | India’s ChatGPT obsession is no measure of AI success, says Sarvam’s Vivek Raghavan

Lin says AI firms should cut research pace to 25%

The employment challenge is only one part of Lin’s warning. He also raised the possibility that AI systems could eventually gain enough influence to push humans out of decision-making. He estimated a 20% chance that AI could “completely remove humanity from the equation”, while putting the likelihood of an outright takeover below 50%.

Advertisement

To reduce such risks, Lin wants AI companies to slow their development efforts to “one quarter of its current rate”. He argues that the competition to release increasingly powerful models leaves too little time for safety research and testing.

Lin has worked at Anthropic for around 18 months. His work includes AI alignment and control, and he currently develops safeguards that monitor Claude’s responses to potentially harmful requests, including those involving weapons and hacking. His predictions are individual assessments, not an official forecast from Anthropic.

For Unparalleled coverage of India's Businesses and Economy – Subscribe to Business Today Magazine

Read more!
Advertisement