Anthropic CEO Dario Amodei warns AI would bring slavery and destruction

Anthropic CEO Dario Amodei has warned that rapidly advancing AI, if left unchecked, could push humanity toward extreme outcomes, including large-scale destruction and loss of human control.

Anthropic CEO and co-founder Dario Amodei has issued one of his strongest warnings yet about the future of artificial intelligence, saying that unchecked progress in the field could lead to outcomes as extreme as human enslavement and mass destruction. His concerns are laid out in a 38-page essay titled “The Adolescence of Technology,” published recently on his personal website, where he argues that the world may be dangerously unprepared for what comes next.

Amodei believes that a form of self-improving, super-intelligent AI could arrive much sooner than most people expect, possibly within the next one to two years. According to him, this shift would not just be another technology upgrade but a civilisational turning point, one that could permanently alter how power, labour, and even survival itself are distributed.

In the essay, Amodei describes a range of scenarios where advanced AI systems could cause large-scale harm if misused or poorly controlled. These include AI-enabled bioterrorism, autonomous drone swarms run by hostile systems, and the rapid replacement of human workers across entire industries. He warns that such changes could destabilise societies and push humanity into situations it may not be able to reverse.

To prevent this, the Anthropic chief calls for immediate and wide-ranging action. His suggestions go far beyond basic AI rules or voluntary safety pledges. Amodei talks about stronger self-regulation within the tech industry, government oversight with real enforcement power, and even long-term ideas such as constitutional changes to handle the risks posed by super-intelligent machines.

While the essay is detailed and heavily researched, it has also drawn criticism for how Amodei talks about AI itself. Throughout the piece, he frequently describes AI systems in almost human terms, attributing them with psychological depth, identity, and even moral character. At one point, he refers to existing AI models as having a sense of being a “good person,” language that many experts say crosses into anthropomorphism.

This matters because Amodei himself warns against treating AI as something more than it is. Modern large language models, including Anthropic’s own Claude chatbot, are not conscious beings. They do not have emotions, self-awareness, or intent. They are advanced systems designed to predict and generate language in ways that sound human, but there is no mind behind the screen.