Anthropic has filed its IPO prospectus, a 335-page document in which roughly 100 pages are devoted to enumerating risks. The filing lays out the company’s particular interpretation of the technology it is building and how it believes that technology will reshape the world.
But Anthropic has been cultivating its public story for years. Its external messaging operates as a deliberate exercise in narrative capture. Earlier this month, Jack Clark, Anthropic’s cofounder and head of policy, told attendees at the International Monetary Fund (IMF) meeting in Bangkok that the singularity has already arrived.
The term “singularity,” introduced around forty years ago by science fiction writer Vernor Vinge, denotes a moment when technology outstrips its creators’ abilities—leaving humans unable to compete with or fully comprehend their own creations. Clark told the assembled economists, bankers and lawmakers: “I believe that this year we have gained convincing evidence that it has now begun.”
Anthropic’s storytelling often paints a vivid and alarming future—borrowed in part from well-worn science-fiction tropes—and frames its technology as both mysterious and dangerous. Simultaneously, the company positions itself as uniquely trustworthy to manage those dangers, a framing that can be attractive to investors.
For example, last year Chris Olah, another Anthropic cofounder, was invited to speak at the Vatican. According to reports, he addressed Pope Leo XIV and other religious scholars and officials about the company’s findings. Olah said: “I will be honest: We keep finding things that are mysterious, even unsettling. We find evidence of introspection. We find internal states that functionally mirror joy, satisfaction, fear, grief, and unease.”
Many researchers would scoff at the implication that Anthropic’s software has humanlike inner experiences. Yet when so much is uncertain about future trajectories, the statements of frontier AI companies are seized upon and treated as lifelines.
I am not asserting that Anthropic’s claims are necessarily incorrect—the question deserves its own investigation. But in the absence of clear, shared knowledge about what is coming, companies like Anthropic are able to fill the explanatory vacuum.
What is Anthropic?
This might sound like a trivial question, but it is surprisingly hard to answer. The question resurfaced in an exchange on X initiated by Nathan Calvin, chief strategy officer at Encode AI, a Washington, D.C.–based advocacy organization for regulation. Calvin criticized a draft bill that Anthropic reportedly circulated confidentially to U.S. lawmakers and that was later leaked. He argued that Anthropic seeks to present itself as the “good actor,” and asked: why should we trust it? “What is Anthropic?” he wrote.
One practical answer: Anthropic is a company whose IPO valuation would exceed $1.5 trillion, enriching its cofounders, including Jack Clark and Chris Olah. But Anthropic has long insisted it is more than a vehicle for private wealth.
Calvin highlighted competing narratives: Anthropic could be viewed as an AI safety organization dedicated to developing powerful but reliable and low-risk technology. Alternatively, it could be seen as a group intent on building superhuman capabilities before anyone else—either because they believe they are the most responsible actors or because of competitive or ego-driven motives.
In practice, Anthropic embodies elements of all of these identities. Dario Amodei, Anthropic’s CEO, replied to Calvin’s post by describing the company’s strategy as a “race to the top”: by demonstrating responsible development, Anthropic aims to raise the industry standard. Amodei acknowledged mistakes in the past and argued that being at the frontier makes the company’s influence on norms more effective.
The tensions of playing two roles
Playing multiple roles is difficult. The priorities of an AI safety organization can be at odds with those of a profit-driven tech company. These tensions surface in concrete policy choices: should Anthropic be allowed to set terms for external safety testers, or should regulators insist on independent access to models? When a piece of legislation appears self-serving, should it be met with skepticism?
The stories Anthropic tells about itself should inform how much trust the public and policymakers grant it. But those narratives reveal a multiheaded entity: the company is many things to many audiences, and that complexity does not automatically translate into greater trustworthiness.
In many respects, a conventional tech company—without the science-fiction framing and dramatics—would be simpler to evaluate. Anthropic’s practice of blending safety rhetoric, rhetorical urgency, and competitive ambition keeps the question open as to whom and what it ultimately serves.
