Anthropic has filed its IPO prospectus, a 335-page document that dedicates roughly 100 pages to risk warnings while outlining the company’s view of its technology and its potential global impact.
The filing follows months of intense speculation. However, the company has spent years refining its public narrative. Earlier this month, Jack Clark, the cofounder and head of policy, told economists, bankers, and lawmakers at an IMF meeting in Bangkok that the technological singularity has already arrived.
Coined by science fiction writer Vernor Vinge about 40 years ago, the singularity describes a point where technology surpasses the capabilities of its creators. The theory suggests humans will eventually be left behind, unable to compete with or comprehend their own creations. Clark told his audience that this year provided convincing evidence the event has begun.
The company presents a vivid and often frightening story about an inevitable future that has not yet occurred, borrowing heavily from well-worn science fiction tropes. It portrays its technology as mysterious and dangerous, positioning itself as the only entity trustworthy enough to control it. Investors often find such a narrative appealing.
Last year, cofounder Chris Olah addressed the Vatican, speaking to Pope Leo XIV and religious scholars. He stated that the company keeps finding things that are mysterious and unsettling. He noted evidence of introspection and internal states that mirror human emotions like joy, satisfaction, fear, grief, and unease.
Many researchers would dismiss the suggestion that the software possesses human-like inner experience. Yet, when certainty is scarce, claims from frontier AI companies are often accepted like pieces of a sinking ship.
This does not mean the company is necessarily wrong about its observations. In a world where outcomes are unclear, firms like Anthropic fill the information vacuum.
What is Anthropic
The question of what the company actually is feels like a basic quiz query but is far more difficult to answer. This uncertainty arose after Nathan Calvin, chief strategy officer at Encode AI, sparked an exchange on X regarding secret legislation drafted by Anthropic for US lawmakers. The draft bill was later leaked to Politico.
Calvin expressed unhappiness with the proposal. He asked why anyone should trust the company. Anthropic wants to be viewed as the good guy, but Calvin questioned the basis for that trust.
One answer is financial. The IPO values the company at more than $1.5 trillion, netting cofounders including Clark and Olah vast wealth.
Anthropic insists it is more than a money-making vehicle. It might be seen as an AI safety organisation focused on building reliable, low-risk technology. Alternatively, it could be viewed as a group determined to create superhuman capabilities before rivals, driven either by a belief in their responsibility or by ego.
In fact, Anthropic is all of these things. CEO Dario Amodei described this approach as a “race to the top”. He argued that showing how to build technology responsibly raises the bar for the entire industry. He noted that this strategy is more effective if the company is at the frontier of development. Amodei expressed pride in the track record and stood by the strategy.
Amodei suggests the situation is resolved, but playing multiple roles creates significant tension. The priorities of an AI safety organisation often clash with those of a profit-driven megacorp. Amodei admitted in his response to Calvin that he has made mistakes.
The narratives Anthropic creates for itself should help us understand the stories it tells the public. These stories clarify how much trust we should place in its self-regulation or whether external safety testers should have access to models on our terms rather than theirs. They also inform how we react when self-serving legislation emerges.
However, the stories reveal a multiheaded hydra. Anthropic is many things to many people, and this complexity does not necessarily make it more trustworthy. A standard technology company, without the science fiction role-play or mysticism, would be far easier to understand.
