Consistently wrong AI alarmists will be wrong again

OPINION:

Artificial intelligence promises great leaps in science, healthcare, agriculture, climate and mathematics. Indeed, India is leveraging AI to create diabetic retinopathy screening tools to catch preventable blindness in patients who would otherwise never see an ophthalmologist. Be My Eyes’ “Be My AI” lets blind users point a phone camera at a scene or document and get a description. AI-driven flood and wildfire prediction gives earlier evacuation windows. AI can catch exploitable flaws in our critical infrastructure running on legacy code before hackers wreak havoc.

Unfortunately, the news inundates you with nothing but AI doom and gloom propagated by AI alarmists.

Last week, former Anthropic researcher Jacob Coxon claimed that the “people building AI earnestly believe that it could kill us all by the end of the decade.” Evan Hubinger, an alignment lead at Anthropic, echoed this sentiment, predicting “it is >10% within the next decade” that “AI could kill all humans.” Even though many experts in artificial intelligence believe this is just not the case, the news ignores this and takes these claims at face value.

Herein lies the rub: The AI alarmists are consistently wrong about AI’s impact.

Just examine their track record.

As Axios reported in May 2025, Anthropic CEO Dario Amodei warned that AI could wipe out half of all entry-level white-collar jobs — and spike unemployment to 10% to 20% in the next one to five years.

A year and a half later, that prediction has not aged well. As The Economist noted, the American labor market is a “far cry from the sort of … mayhem prophesied by some.”

The Associated Press recently reported that “layoffs remain at historically low levels.”

Indeed, the unemployment rate stands at 4.1% as of August 2026, per the Bureau of Labor Statistics — near the level at which Mr. Amodei made his prediction and far below his projected range of 10% to 20%.

Mr. Amodei made his most alarming prediction in 2023, when he testified before the Senate that humanity faced “medium-term risks … in two to three years” that AI could become capable of enabling large-scale biological destruction. It has been more than three years, and thankfully, this has not come to pass either.

Job-specific predictions about software developers have not materialized either. In March 2025, at a Council on Foreign Relations event, Mr. Amodei said, “I think we will be there in three to six months, where AI is writing 90% of the code. And then, in 12 months, we may be in a world where AI is writing essentially all of the code.”

More than a year later, The Wall Street Journal reported, “That hasn’t happened. It isn’t likely to happen any time soon, either.” It would seem that Mr. Amodei is even getting it wrong when it comes to a job — coding — he should presumably be very familiar with as a tech CEO.

Even Nobel laureates are not immune. In 2016, Geoffrey Hinton told a machine learning conference in Toronto that people should stop training radiologists because, within five years, machine learning would do the job better. The deadline came and went. Radiology is now one of medicine’s most in-demand specialties, and not because the technology failed to arrive: It is the largest category of FDA-cleared AI devices in American medicine. Mr. Hinton was right about the machines and wrong about the people. He conceded as much to The New York Times in 2025, and he remains among the most-quoted voices on what AI will do next.

Other policy advocates of so-called effective altruism have delivered similar dire warnings advocating for bans on advanced artificial intelligence models. As tech researcher Nirit Weiss-Blatt observed last year, “Humanity now lives with … models exceeding the doomers’ nightmarish threshold.” A year later, this too has not borne out. Advanced models built on the technologies warned about have since been deployed without catastrophic or existential consequences.

Skeptics of artificial intelligence are justifiably concerned about how the technology could be abused or worse. Its impact on children, autonomous hacking and the market becoming increasingly concentrated among a few companies are all legitimate fears.

Yet these concerns can be addressed by targeted, incremental policy changes. There is no reason at this juncture to kill AI in its cradle, especially when these fears are amplified by the hype generated by opportunistic AI companies. Surely, the federal government and industry should take measures to protect consumers and humanity in general.

Although artificial intelligence companies must build responsibly, regulatory advocates have an equal responsibility not to needlessly panic the public about the dangers of AI.

• Joel Thayer is a senior fellow for artificial intelligence and emerging technology policy at the America First Policy Institute. He also serves as president of the Digital Progress Institute and as a tech and telecom lawyer.