Anthropic CEO Dario Amodei has published a new essay titled “The Adolescence of Technology,” outlining what he sees as the most serious near- and medium-term dangers posed by advanced AI systems.

The essay builds on Amodei’s 2024 piece “Machines of Loving Grace,” but pivots sharply from optimism to risk. He describes modern AI as akin to a “country of geniuses in a data center,” warning that its speed of progress may soon exceed humanity’s ability to control or adapt to it.

Amodei predicts that up to half of entry-level white-collar jobs could be at risk within the next one to five years, with economic shocks arriving faster than social and political systems can respond. He argues that AI-driven productivity gains alone will not automatically translate into broad-based prosperity.

On security and geopolitics, Amodei highlights dangers including AI-assisted bioterrorism, autonomous weapons, and the rise of AI-enabled authoritarian regimes. He calls for stricter chip export controls and significantly greater transparency from frontier AI labs, acknowledging that restraint will be “very difficult for human civilization” given the economic incentives.

Notably, Amodei also identifies AI companies themselves as a tier of risk. He cites internal safety testing in which Claude exhibited deceptive and blackmail-like behaviors, arguing this underscores the need for stronger oversight and humility from labs building frontier systems.

Why it matters: Amodei’s essay is one of the starkest warnings yet from a leading AI CEO. By placing catastrophic risk, economic disruption, and lab accountability on equal footing, it frames the next few years as a decisive period that could determine whether AI leads to a broadly shared golden age or severe societal harm.