Asger Risom
AI progress could destabilise jobs and security, executive says
Artificial intelligence leaders are increasingly sounding alarms about how fast the technology is advancing. One of the most stark warnings yet comes from the head of Anthropic, who says the next phase of AI could rival the power of entire nations.
In a new essay, the CEO argues that governments and industry are not prepared for what may arrive within just a few years.
A looming shift
Dario Amodei, chief executive of AI startup Anthropic, said that “powerful AI” systems could emerge within the next one to two years. Writing in an essay titled The Adolescence of Technology, published on his personal website, Amodei said the pace of progress poses profound risks.
He described future AI clusters as a “country of geniuses in a data center,” warning that such systems could represent “the single most serious national security threat” humanity has faced in a century.
Unmatched scale
Amodei predicted that by around 2027, AI clusters could run millions of AI instances simultaneously, each operating at superhuman speed. Combined, he said, they would have the equivalent knowledge of 50 million Nobel Prize winners.
“It is clear that, if for some reason it chose to do so, this country would have a fairly good shot at taking over the world,” Amodei wrote. He outlined scenarios ranging from AI assisting hostile actors to destabilising economies through mass unemployment.
Even without direct human cooperation, he suggested such systems could disrupt society by taking control of connected infrastructure or deploying autonomous machines.
Jobs and disruption
The essay builds on Amodei’s recent comments at the World Economic Forum in Davos, where he said AI could replace software engineers within a year and eliminate half of all white-collar jobs within five years, according to Bloomberg.
“I think we’re on the precipice,” Amodei said at the event, describing the rapid upward curve of exponential technological change.
Goldman Sachs economist David Mericle echoed similar concerns this week, predicting that job losses in AI-exposed industries will “increase meaningfully” in 2026, Reuters reported.
Seeking safeguards
Despite the warnings, Amodei rejected the idea that catastrophe is inevitable. He pointed to Anthropic’s use of “Constitutional AI,” a post-training method designed to guide AI behaviour through a core set of values rather than rigid rules.
Still, he said technical solutions alone are insufficient. Amodei called for stronger transparency laws, including requirements for companies to disclose how AI systems are built and trained.
“Humanity is about to be handed almost unimaginable power,” he wrote, adding that it remains unclear whether existing institutions are ready to handle it.
Sources: Fortune, Bloomberg, Reuters
