From anti-war physicist to AI power broker, Anthropic CEO Amodei takes his safety approach public
AI generated
asserted
Amodei → take → approach
Long before Anthropic CEO Dario Amodei butted heads with the Pentagon over whether the US military could use Anthropic’s models for mass domestic surveillance and fully autonomous weapons, he was fighting another battle.
uncertain
he → butt → battle
At the California Institute of Technology in 2003, Amodei called on his classmates to join him in protesting the US invasion of Iraq, writing in an op-ed that citizen-scientists should not be “reduced to amoral technicians who run the machines of war”.
asserted
who → call → war
On Nov 19, as 43-year-old Amodei led his startup to an IPO that raised a record US$81 billion (S$104.61 billion) and valued it at about US$1.5 trillion, the listing cemented his transformation from an idealistic physics student into one of the most powerful custodians of the AI age.
asserted
listing → lead → age
“We will not bend to either business nor political pressure,” Amodei said at the Nasdaq on Nov 19 in New York.
asserted
Amodei → bend → York
“Today puts the company in a stronger financial position, and in doing so only strengthens our resolve to do the right thing.”
asserted
Today → put → thing
Amodei has spent much of his ascent emphasising curbs on a technology whose extraordinary potential he has devoted his career to unlocking, and has now made his company one of the most valuable on Wall Street.
asserted
company → spend → Street
Even as Anthropic prepared for its debut, Amodei penned a nearly 4,000-word essay in September, warning how AI’s advances could outpace human control and urging a slowdown in development.
uncertain
advances → prepare → development
Days later, Anthropic released a new version of one of its most powerful models in a challenge to rival OpenAI, where he worked before quitting over differences related to AI safety.
asserted
he → release → safety
In that essay, Amodei denied any contradiction between his quest for safe AI and Anthropic’s race to develop the most powerful models.
asserted
Amodei → deny → models
“I have grappled with this duality of risk and benefit since the beginning of Anthropic,” he wrote.
asserted
he → grapple → Anthropic
“Not building the technology deprives humanity of benefits or simply places AI in the hands of authoritarian powers, while building it too fast is reckless.”
asserted
building → build → it
Anthropic was a commercially viable company spurring a safety “race to the top”, he said.
asserted
he → spur → top
This approach has drawn criticism both from advocates of greater safety measures and from Silicon Valley executives who have warned against slowing down innovation, such as Nvidia’s former CEO Jensen Huang.
asserted
who → draw → Huang
Amodei’s determination to impose some limits on AI also earned him enemies in Washington, culminating in an extraordinary showdown with the Trump administration early in 2026.
asserted
determination → impose → 2026
Following a meeting between Amodei and US Defence Secretary Pete Hegseth, the government set a deadline for Anthropic to accept the military’s contract terms, which the company rejected.
asserted
company → follow → which
That led to a government blacklist of its technology that the Claude maker challenged in court.
asserted
maker → lead → court
Anthropic declined to comment for this story.
asserted
Anthropic → decline → story
Anthropic shares rose sharply on their first day of trading, debuting at US$111.50, 34% higher than the US$83 IPO price.
asserted
shares → rise → price
The price of caution
Though the IPO was expected to mint thousands of paper millionaires, Amodei has said the company's safety focus has come at a cost.
asserted
focus → expect → cost
In October, he responded to a post on X that challenged Anthropic’s safety record and criticised a change to its pledge around halting AI development if it cannot minimise risks.
asserted
it → respond → risks
Amodei said the company had “likely sacrificed a significant chunk of a several-month model lead” by delaying the release of its powerful Mythos models over hacking risks and initially limiting access to cybersecurity defenders.
asserted
company → say → defenders
Anthropic is spending billions of dollars on computing capacity each year, making it harder as a publicly traded company – albeit one where its seven co-founders or a trust wield voting control on key matters – to justify slowing model development as shareholders seek returns.
asserted
shareholders → spend → returns
In its IPO filing, though, the company noted that its corporate setup could lead to decisions that “may conflict with short-, medium-, or long-term financial interests and business performance”, and said it had paused training of a next-generation flagship AI model for more than a month over security concerns.
uncertain
it → note → concerns
Dario Amodei and his sister and co-founder Daniela grew up in San Francisco’s Mission District, just a few miles from the spot where Anthropic would eventually occupy a glass tower with Silicon Valley perks such as a gourmet espresso bar.
asserted
Anthropic → grow → bar
He was voted “most likely to succeed” by his eighth-grade classmates, a yearbook snippet shows and admitted to San Francisco’s selective Lowell public high school.
asserted
snippet → vote → school
He went to the California Institute of Technology before transferring to Stanford, and then attended Princeton, where he earned a PhD focusing on biophysics and computational neuroscience.
asserted
he → go → biophysics
After his father’s death from hepatitis C about two decades ago – a few years before medical advances dramatically improved survival rates – Amodei pursued artificial intelligence that could accelerate medicine.
uncertain
that → improve → medicine
In a podcast, talking about his dad’s illness, Amodei pushed back against criticism that he was simply an AI “doomer”.
asserted
he → talk → criticism
Amodei’s preoccupation with AI safety has influenced his management style.
asserted
preoccupation → influence → style
Daniela Amodei has oversight of day-to-day operations at the company, while Dario Amodei devotes almost 40% of his time to framing the culture and vision of Anthropic through essays and fortnightly company-wide meetings that employees have dubbed “Dario Vision Quests”.
asserted
employees → have → that
“Operationally, the buck stops with her," a former employee said of Anthropic President Daniela Amodei.
asserted
employee → stop → Amodei
“Strategically, it stops with him.”
asserted
it → stop → him
Dario Amodei said in a Bloomberg interview published in June that he had only one direct report, calling the structure “incredibly freeing”.
One of his favourite books is The Making Of The Atomic Bomb, written by Richard Rhodes.
asserted
One → say → Rhodes
Talking about Robert Oppenheimer, the bomb’s architect, Amodei said: “In some ways I actually see Oppenheimer as a failure case, as what should not happen.”
asserted
what → talk → case