Anthropic co-founder Daniela Amodei and her husband kept a “personal advisory council” of stuffed animals to work through workplace problems, a habit that started at OpenAI and now sits beside a hard push for AI rules critics call market capture.
Breitbart News reported that an Atlantic account, adapted from journalist Kevin Roose’s forthcoming book, describes how Amodei and husband Holden Karnofsky assigned character profiles to plush toys and used them on office dilemmas.
The practice began while both were still at OpenAI, years before Anthropic existed. Few colleagues knew. Those who did were never sure how seriously the pair treated the “advice.”
In 2023, Amodei told Roose the toys helped her analyze different personality types. The council had names and roles: Jinji, a lazy cat; Maura, a pink bear cast as type-A; and Beary Bonds, a panda tied to generosity and empathy.
Daniela Amodei joined OpenAI in 2018 after studying English literature. OpenAI co-founder Greg Brockman had given a toast at her 2017 wedding to Karnofsky, who later sat on OpenAI’s nonprofit board.
Her reputation inside the company was as the more emotionally attuned Amodei sibling. Responsibilities grew until she was managing technical teams. Friction followed with Brockman and then-executive Mira Murati.
Her brother Dario, now Anthropic’s CEO, grew distrustful of Sam Altman and Brockman. Reporting on the period describes a “Dario’s clique,” private Slack channels, and separate safety meetings away from the main leadership track.
By 2020, Dario was recruiting people to leave and build a new company around an AI-safety mission. Daniela took the lead in negotiating the split with OpenAI, staff nicknamed it “the divorce.”
In December 2020, the siblings left and founded Anthropic with Jack Clark, Sam McCandlish, Tom Brown, Chris Olah, and Jared Kaplan. Ben Mann joined later as the eighth co-founder. The shop started as a roughly 50-person safety lab. It has since grown to thousands of employees. Daniela is president and board chair, with oversight across research, engineering, product, governance, and commercial operations. Karnofsky later joined Anthropic’s technical staff. The company has been described as eyeing a $2 trillion IPO.
Anthropic has stood at the center of calls for an organized slowdown in AI development and for government regulation of the technology. The same industry that ships powerful models also asks Washington to write the rules. China’s role in the race adds another layer of pressure and complexity.
David Sacks has pointed out that when AI titans such as Dario Amodei, Sam Altman, and Elon Musk urge a slowdown, the move is not pure altruism. The incentive, in that reading, is to lock in position and wall the sector off from antitrust exposure, a conflict President Donald Trump recognizes.
Breitbart News editor-in-chief Alex Marlow has been blunter about motive and structure. In recent remarks, Marlow cast the regulatory push as power politics dressed up as public virtue.
Marlow said:
“They are building a cartel right in front of your face, and they’re pretending like they’re the altruist in the process.”
He tied the Amodei orbit to a Berkeley-based group called METR. Marlow said Amodei proposed the group and noted that one of the most recognizable names connected to it is Holden Karnofsky, married to Daniela Amodei, co-founder of Anthropic with her brother.
Marlow continued:
“So, Dario’s solution to the deadly AI that he created is to appoint a group tied to his own family.”
And:
“Dario Amodei wants an AI regulator controlled by Dario Amodei and Dario’s sister, so they can create the framework under which all of us operate.”
On the reason for the rules push, Marlow said these executives are after “power and market capture.” He added they are “not interested in saving humanity,” and that they want to “use AI to be able to leverage market capture and, God forbid, create these international bodies that will control more of your lives.”
Roose’s Atlantic piece, drawn from his book The AGI Chronicles: The Inside Story of the Race to Create an Artificial Superintelligence, puts the stuffed-animal habit in the same career arc as the OpenAI exit and Anthropic’s rise. Amodei’s 2023 explanation framed the toys as a tool for reading personality types. The public record in this reporting does not show a formal governance role for Jinji, Maura, or Beary Bonds. It does show founders who left a rival lab citing safety distrust, built a major AI firm, and now press for the slowdown and the rulebook.
Taxpayers and voters are being asked to accept sweeping AI controls on trust. The people asking built the systems, staffed the labs, and, in this account, once ran live workplace problems past a cat, a pink bear, and a panda with assigned vibes.
When the same circle that profits from the models also writes the case for the referee, common sense says read the incentives before you hand them the whistle.
