Back to Flash News

Anthropic researcher Tao Lin says AI may leave nearly all humans permanently unemployed

Anthropic safeguard team researcher Tao Lin said in an interview released by Palisade Research on November 15 that he believes it is "very likely" that nearly all humans on Earth will permanently lose their ability to work and will need to rely strictly on government welfare to make a living, and that human employment will disappear. He expects "broad superhuman AI" (reaching or exceeding the level of the best human experts in almost every field) to emerge in about 1.3 years, around the end of 2027.

Lin also pointed out that this will not immediately lead to mass unemployment. He expects a gradual transition: people first lose their current jobs and move into temporary work, partly because there are not enough robots yet. He clearly answered no to whether this would happen within the next 15 months. Lin said these were personal predictions, not Anthropic’s official stance, and said that people should be angry at him and everyone developing AI.

Original source:

  1. Title: Anthropic Researcher: AI Will Leave Most Humans Unemployable
    Publisher: Unusual Whales
    Publication date: 2026-11-20
    URL: unusualwhales.com/news/anthropic-researcher-ai-leave-humans-unemployable-welfare
    Source type: media report (based on a Palisade Research interview)
    Verification status: this reading covered the relevant report and interview transcript excerpt, verification date 2026-11-20
    Supporting content: supports Lin’s reported speech, the superhuman AI timeline, the reliance-on-welfare statement, the gradual transition, and the qualification of no within 15 months.

ABAB AI Insight

Tao Lin currently works on Anthropic’s safeguard team and previously participated in research on automated alignment and AI control, and has been at the company for about a year and a half. His interview is part of Palisade Research’s "From Inside AI" series, which aims to record personal views of employees at AI companies rather than company policy. Anthropic’s own public data such as economic indices previously showed that the proportion of augmented use in sampled conversations was higher than automation, but that was a usage snapshot rather than an employment forecast.

On capital and deployment pathways, Lin attributes the gap between capability attainment and actual replacement to an insufficient number of physical robots. Cognitive tasks that software can handle may be affected earlier, while jobs that depend on physical execution require factories and large-scale robotics; he mentioned that billions of units might be needed. This distinction is consistent with historical patterns of automation in which cognitive tasks were replaced before physical tasks, but it does not constitute a forecast of unemployment rates in any specific industry.

On policy implications, Lin suggested slowing the pace of AI research to about one quarter of the current rate so that safety work can keep up. He also gave an estimated probability of about 20%, saying that advanced AI might eventually marginalize humanity. These numbers and timelines are all his personal judgments, without independent model or company endorsement. A comparable case is public interviews by employees of other frontier labs, in which they make differentiated statements about capability timelines and employment impacts, typically emphasizing the separation between personal views and company positions.

ABAB News · Cognitive Laws

  1. There is a deployment lag between capability reaching superhuman levels and large-scale job replacement, especially constrained by the physical scale of robots.

  2. The "permanent unemployment" wording in personal predictions must be separated from an explicitly denied short-term time window so as not to equate the capability timeline with the employment timeline.

  3. Public predictions by people inside AI companies are personal views and do not automatically constitute company policy or an auditable employment forecast.