Dario Amodei
Co-founder & CEO, Anthropic
A 25% chance of catastrophe is worth racing for the other 75: build the safest frontier AI, warn loudly, and pace the industry before it paces us.
Co-founder and CEO of Anthropic and the AI debate's designated honest man: the executive who publishes essays on AI curing disease while warning of a "white-collar bloodbath," puts his p(doom) at 25%, and in July 2026 joined the call to deliberately pace frontier development — while shipping frontier models. His disagreements with Sam Altman and Jensen Huang map the industry's real fault lines.
Basis for the leaning assessment
Avoids party politics; his safety-forward regulatory stances and labor-disruption warnings read center-left, and he has criticized both parties' AI postures. The vote will hinge on how people read "safety."
Current Stances
Stances are editorial interpretations, always shown with the statements and actions they rest on.
Transparency requirements now — and pace the frontier
Opposed the state-regulation moratorium, proposed federal transparency standards, and backed the 2026 pacing petition: regulation as safety engineering, not obstruction.
Supporting statements & actions (2)
In a New York Times op-ed, opposed the proposed 10-year federal moratorium on state AI regulation, calling instead for a national transparency standard for frontier developers.
A rare case of a frontier-lab CEO lobbying for more regulation, not less — "too blunt an instrument," he wrote of the moratorium.
Anthropic joined OpenAI in backing a petition calling on the US government to support deliberately pacing frontier AI development, so society can adapt to each new capability level.
The pacing framework Amodei had advocated became, briefly, the industry consensus position.
A white-collar bloodbath is coming
Half of entry-level white-collar jobs could vanish within five years; he says the industry owes the public honesty about it. The direct counterpoint to Huang and Zuckerberg.
Supporting statements & actions (1)
AI could wipe out half of all entry-level white-collar jobs — and spike unemployment to 10-20% in the next one to five years. Lawmakers and CEOs need to stop "sugarcoating" what is coming.
The "white-collar bloodbath" interview — the most-cited labor warning of the AI boom, made about his own industry's product.
The "white-collar bloodbath" interview dominated LinkedIn and X for weeks and entered the jobs-debate lexicon.
Powerful AI is close — and worth the risk if done right
A 25% chance of catastrophe, a 75% chance of a compressed century of progress: he builds frontier AI because he thinks the odds justify it, and says so with unusual precision.
Supporting statements & actions (2)
In "Machines of Loving Grace," argued powerful AI could compress a century of scientific progress into 5-10 years — curing most diseases and lifting billions from poverty — "if we get it right."
His definitive optimist case, written explicitly to balance his reputation as a doomer.
There's a 25% chance that things go really, really badly — and a 75% chance that things go really, really well.
His answer when asked for his "p(doom)" — the probability that AI catastrophically harms humanity.
The "25%" clip became the most-quoted risk estimate in AI — invoked by both sides of every safety argument.
Goals
Short-term
Get government to pace the industry
ActiveTransparency standards plus the 2026 pacing petition: slow everyone symmetrically, so safety isn't a competitive disadvantage.
Declared: 2026-07-28Source ↗
Long-term
Reach powerful AI safely — and first among the careful
ActiveAnthropic's wager: the 75% good outcome is won by being at the frontier, not by abstaining from it.
Declared: 2024-10-11Source ↗
Consistency Tracker
Does each new statement or decision square with the record? Verdicts are not accusations: both sides of the record are presented so readers can judge for themselves.
Earlier statement / action
In "Machines of Loving Grace," argued powerful AI could compress a century of scientific progress into 5-10 years — curing most diseases and lifting billions from poverty — "if we get it right."
His definitive optimist case, written explicitly to balance his reputation as a doomer.
Later statement / action
AI could wipe out half of all entry-level white-collar jobs — and spike unemployment to 10-20% in the next one to five years. Lawmakers and CEOs need to stop "sugarcoating" what is coming.
The "white-collar bloodbath" interview — the most-cited labor warning of the AI boom, made about his own industry's product.
The "white-collar bloodbath" interview dominated LinkedIn and X for weeks and entered the jobs-debate lexicon.
The essay promising a compressed century of progress and the interview predicting a white-collar bloodbath look like whiplash but are one worldview: the same capability that cures diseases eliminates entry-level work. He states both halves with equal volume — which is precisely what he accuses the rest of the industry of refusing to do.
Earlier statement / action
There's a 25% chance that things go really, really badly — and a 75% chance that things go really, really well.
His answer when asked for his "p(doom)" — the probability that AI catastrophically harms humanity.
The "25%" clip became the most-quoted risk estimate in AI — invoked by both sides of every safety argument.
Later statement / action
Anthropic joined OpenAI in backing a petition calling on the US government to support deliberately pacing frontier AI development, so society can adapt to each new capability level.
The pacing framework Amodei had advocated became, briefly, the industry consensus position.
A 25% catastrophe estimate followed by a petition to pace the entire frontier is belief followed by matching action. The standing critique — that Anthropic keeps shipping frontier models while asking everyone to slow down — is answered in his telling by symmetry: pace everyone, including us. The record here supports the claim that the warnings and the conduct point the same way.
Most-Cited Sources from Voterscommunity reading list
Articles readers attached to their poll votes as evidence. Only pages cited independently by two or more voters are listed.
- 1Liberal 1Conservative 3
- 2www.axios.com/2025/09/17/anthropic-dario-amodei-p-doom-25-percent ↗
Cited by 4 voters
Liberal 3Conservative 1 - 3Liberal 2Conservative 2
Most-Discussed Moments
The statements and actions that set social media on fire — top 2, in chronological order. Every entry links to its source.
- 2025-05-28Interview
AI could wipe out half of all entry-level white-collar jobs — and spike unemployment to 10-20% in the next one to five years. Lawmakers and CEOs need to stop "sugarcoating" what is coming.
The "white-collar bloodbath" interview — the most-cited labor warning of the AI boom, made about his own industry's product.
The "white-collar bloodbath" interview dominated LinkedIn and X for weeks and entered the jobs-debate lexicon.
- 2025-09-17Speech
There's a 25% chance that things go really, really badly — and a 75% chance that things go really, really well.
His answer when asked for his "p(doom)" — the probability that AI catastrophically harms humanity.
The "25%" clip became the most-quoted risk estimate in AI — invoked by both sides of every safety argument.
Timeline of Statements & Actions5 records, newest first
Anthropic joined OpenAI in backing a petition calling on the US government to support deliberately pacing frontier AI development, so society can adapt to each new capability level.
The pacing framework Amodei had advocated became, briefly, the industry consensus position.
There's a 25% chance that things go really, really badly — and a 75% chance that things go really, really well.
His answer when asked for his "p(doom)" — the probability that AI catastrophically harms humanity.
The "25%" clip became the most-quoted risk estimate in AI — invoked by both sides of every safety argument.
In a New York Times op-ed, opposed the proposed 10-year federal moratorium on state AI regulation, calling instead for a national transparency standard for frontier developers.
A rare case of a frontier-lab CEO lobbying for more regulation, not less — "too blunt an instrument," he wrote of the moratorium.
AI could wipe out half of all entry-level white-collar jobs — and spike unemployment to 10-20% in the next one to five years. Lawmakers and CEOs need to stop "sugarcoating" what is coming.
The "white-collar bloodbath" interview — the most-cited labor warning of the AI boom, made about his own industry's product.
The "white-collar bloodbath" interview dominated LinkedIn and X for weeks and entered the jobs-debate lexicon.
In "Machines of Loving Grace," argued powerful AI could compress a century of scientific progress into 5-10 years — curing most diseases and lifting billions from poverty — "if we get it right."
His definitive optimist case, written explicitly to balance his reputation as a doomer.