A new profile from The Information, cited by MarsBit, portrays Anthropic CEO Dario Amodei as a deeply split figure in the AI race: a founder who repeatedly warns that artificial intelligence could wipe out humanity, while continuing to accelerate the push toward more capable systems.
The article, written by Cory Weinberg after 42 days of reporting and interviews with subordinates, competitors and investors, frames Amodei’s public and private conduct around that same tension. He is presented as both an evangelist for AI safety and a central builder of the technology he believes could become an existential threat.
A long-running fixation on AI safety
According to the report, Amodei’s focus on AI risk goes back at least a decade. In 2017, he was already warning about the safety dangers that artificial general intelligence, or AGI, might bring. By 2018, as OpenAI’s early model progress was drawing excitement, he had adopted what the story describes as an unusually rigid posture on internal safeguards.
One example involved a rule that all core OpenAI research had to undergo strict “information hazard” review before publication, with the aim of preventing technical details from leaking to competitors. The profile says that level of caution left some colleagues feeling stifled.
It also recounts a 2019 episode, when Microsoft was preparing a $1 billion investment in OpenAI as the startup faced financial pressure. Amodei’s safety team, the article says, forced a pause of several months in those crucial fundraising talks because of concern that GPT-3, still in training at the time, could turn into an uncontrollable AGI.
The Information report adds that Amodei’s views on security shaped even his personal workflow. In his San Francisco apartment, he reportedly refused to use computers connected to the public internet. Sensitive memos were written on an air-gapped machine, then printed and physically carried to the office rather than sent by email or uploaded to the cloud.
That bunker-like mentality, as the piece describes it, peaked in 2020. Amodei later left OpenAI with a group of colleagues and founded Anthropic. Inside that founding circle, the report says, Sam Altman’s drive toward rapid commercialization was seen as a dangerous path, and that unease evolved into a distinct internal culture at Anthropic.
Rejecting softer messaging on AI
The profile says that earlier this year, when public sentiment around AI had turned sharply negative, some Anthropic board participants proposed a charm offensive. The suggestion was to lean on the company’s industry influence and emphasize AI’s possible benefits in areas such as anti-cancer drug discovery and the treatment of rare diseases.
Amodei rejected that proposal on the spot, according to the report. His view was that this was not a normal public-relations problem but a real “existential risk.” The company ultimately moved in the opposite direction. During the World Cup, Anthropic ran ads with dark and oppressive imagery, including a row of gravestones, to signal the possibility that AI could kill everyone.
Investors and employees describe a different kind of CEO
The same report says Anthropic investors privately share a broad view that Amodei feels “more like a religious leader than a CEO.” He is described as largely absent from routine management tasks such as compensation, organizational structure and staffing changes. Those responsibilities are said to fall mostly to his sister, Daniela.
Amodei’s own attention is directed elsewhere. In one recent example, he listened to Anthropic’s internal economists present rough models on what a “singularity” could mean. A slow singularity, they estimated, might add 0.5 percentage points to U.S. economic growth with employment largely stable. A fast singularity could add 4 percentage points while sharply increasing unemployment. After hearing the presentation, he reportedly said it matched what he already had in mind.
Employees quoted in the story speak of him in near-mystical terms. One says he can see the future. Another says “the singularity” is in his head. His twice-monthly all-hands meetings are referred to internally as the “Dario Amodei vision tour,” where he works through a sprawling document and shares blunt views on AI, politics, war and society.
The report says his direct, peer-like communication style on Slack has helped build intense loyalty inside the company. In a Silicon Valley environment where many executives are trained as growth operators, the profile presents Amodei as someone whose authority rests more on doctrine and mission than on standard corporate management.
Anthropic’s rise and investor unease
The profile pairs that leadership style with extraordinary business growth. Anthropic, founded five years ago, now has just over 3,000 employees and a reported valuation of $965 billion. Using a roughly $4.1 billion valuation from May 2023 as the baseline cited in the article, that would amount to an increase of about 235 times in three years.
The story also compares Anthropic with Oracle. As of the end of May 2026, Oracle had about 141,000 employees, while Anthropic’s workforce was only 2% to 3% of that size. Even so, the company is described as posting one of the fastest revenue growth trajectories on record.
Yet the article says some investors are unsettled rather than reassured. One veteran Silicon Valley investor who studied Anthropic early wrote in his notes that his biggest fear was that Amodei simply did not care about money. Another public-fund manager preparing to make a large bet on a future IPO reportedly came away from a meeting pleading for him to stop frightening everyone.
The central contradiction
The piece leaves readers with a stark paradox. Amodei’s position, as presented in the report, is that the best way to prevent AI from destroying the world is to personally help build the most powerful AI before someone else does. If the train toward the singularity cannot be stopped, he appears to believe it is safer for him to hold the controls than to leave them to people he sees as less cautious.
That contradiction now sits at the center of Anthropic’s identity. The company presents itself as safety-focused, but it is also one of the most important competitors in the race to build advanced AI. The profile does not resolve that tension. It shows, instead, that Amodei’s warnings about existential risk have never pulled him out of the competition itself.
The source material referenced in the MarsBit item includes The Information article and an episode of the podcast Listen to the Story. The MarsBit page also says the Chinese article originally came from the WeChat account “新智元,” with “ASI启示录” listed as author and “大卫 桃子” as editors.