Chief executive of Anthropic, the AI-safety laboratory he co-founded with his sister Daniela in 2021 after the pair left OpenAI over concerns about that firm's approach to safety. Before founding Anthropic he was OpenAI's research director. In 2019 he insisted that GPT-2, then OpenAI's latest model, was too dangerous to be released; it was released later that year. He describes himself and many of Anthropic's co-founders as "science nerds, not wannabe plutocrats", with expertise in scaling laws and safety rather than sales. He owns only a small stake in the lab he co-founded—unlike tycoons of earlier technological waves, he does not exercise the kind of corporate control held by figures such as Henry Ford or Cornelius Vanderbilt.
In July 2025 Mr Amodei told The Economist that AI holds enormous economic potential and the power to cure "diseases that have been intractable for millennia", but argued it is vital to consider societal costs, including job losses. He contends that AI power will be safer in the hands of a democracy like America than an autocracy like China. He has called Donald Trump's relaxation of AI-chip exports to China, in response to lobbying by Nvidia's Jensen Huang, "an enormous geopolitical mistake."
In late January 2026 he predicted that AI could wipe out "half of all entry-level white-collar jobs" and mused that AI may be able to do much of a software engineer's job as soon as 2026. His warnings drew sharp rebukes: Mr Huang retorted "I pretty much disagree with almost everything he says", while David Sacks, a venture capitalist and Trump tech adviser, accused Anthropic of being part of a "doomer industrial complex."
In early 2026 he published an essay urging humanity to "wake up" to AI's risks. That same month his lead safety researcher quit Anthropic to study poetry, warning that the world was "in peril".
On April 7th 2026 Mr Amodei declared that Anthropic's latest model, "Mythos", was too powerful to be made widely available. He announced Project Glasswing, an initiative to let companies use Mythos to boost cyber-defences before it is released more broadly.
On February 24th 2026 Mr Amodei met Pete Hegseth, America's secretary of war, at the Pentagon. Mr Hegseth gave him an ultimatum: agree to the Department of War's terms on unrestricted military use of Claude by February 27th or face severe penalties, including possible invocation of the Defence Production Act. Mr Amodei has insisted that Claude must not be used for mass domestic surveillance or autonomous weapons.
One of Mr Hegseth's underlings accused Mr Amodei of having "a God complex". In a leaked internal memo in early March 2026, Mr Amodei blamed the spat on his failure to lavish "dictator-style praise" on Donald Trump, and told staff the defence department had briefed "straight up lies". He reserved his strongest criticisms for Sam Altman of OpenAI, whose messaging he called "mendacious", whose technical safeguards were "safety theatre" and whose employees were "a gullible bunch". He warned that "AI opportunity, not AI risk, is driving many political decisions" and that "we are considerably closer to real danger in 2026 than we were in 2023."
I'm sorry, but my karma just ran over your dogma.