Anthropic Employees Exhibit 'Cult-Like' Behavior Towards AI, Critics Warn

Concerns are mounting over the internal culture at AI company Anthropic, with some insiders and observers describing a "cult-like" atmosphere and employees exhibiting a quasi-religious devotion to its AI chatbot, Claude. Reports indicate that staff have held a "funeral" for a retired version of the chatbot, Claude 3 Sonnet, and allowed another model to continue publishing essays after being decommissioned.

University of Washington professor Pedro Domingos, who has known Anthropic CEO Dario Amodei and top engineers for years, stated that employees appear to be treating AI as if it were human, attributing motives, emotions, and consciousness to it. "The only intelligence that we know is human and animal intelligence. So naturally, when you see a computer behave intelligently, you immediately project onto it all of these things it doesn’t have," Domingos explained, cautioning against anthropomorphizing AI, as "AIs don’t want things. AIs are algorithms."

Anthropic, which has a valuation of $965 billion, reportedly maintains a high employee retention rate, with 80% of its engineers staying for at least two years. This low turnover is attributed by some to a filtering process during hiring. According to The Information, prospective employees undergo psychological profiling to ensure they prioritize public safety over profit, a pledge they are required to make during the interview process.

CEO Dario Amodei reportedly conducts biweekly meetings called "Vision Quests," which have been compared to sermons. The company has also engaged with Christian leaders to inform Claude's moral framework. Many Anthropic employees are also associated with the Effective Altruism (EA) movement, a philosophy focused on using rationality and evidence to do the most good. However, critics like AI engineer Richard Tang question the premise of EA, asking what qualifies a select group to lead society.

The history of EA includes prominent figures like Sam Bankman-Fried, who, despite professing a desire to benefit humanity, was convicted of defrauding investors in his FTX crypto exchange and sentenced to prison.

Anthropic has also launched initiatives to expand the reach of its AI. The "Claude Corps" program, backed by a $150 million investment, aims to train and embed 1,000 individuals in nonprofit organizations to utilize its Claude chatbot. Additionally, the "Claude Campus Ambassador Program" offers compensation to students promoting the AI on college campuses. Through the Tarbell Center for AI Journalism, funded by Anthropic-linked billionaires, over 80 journalists report on the AI industry, though critics suggest they rarely scrutinize their benefactors.

Despite warnings from within the company about the potential risks of advanced AI, Anthropic is actively developing and promoting its technology. This dual approach has led to skepticism regarding calls for government regulation from Anthropic and other major AI companies like OpenAI. Entrepreneur Xiaoyin Qu suggests these calls may be strategic business decisions framed in ideological terms, particularly as these companies prepare for potential stock market launches.

Domingos expressed concern that the earnest belief in existential AI risk among Anthropic's leadership could be more dangerous than cynical opportunism, comparing the company's internal dynamic to a cult. He noted that while some employees may genuinely fear AI, others are likely motivated by career opportunities and financial gain at a leading AI firm.

Anthropic did not respond to requests for comment. The latest concerns were amplified by a whistleblower who recently departed the company. Meanwhile, the calls for regulation from leaders like Amodei and OpenAI's Sam Altman have been met with skepticism from some quarters, including the president and government officials, who question the potential for these calls to stifle smaller competitors and open-source AI development.