Anthropic Unveils IPO Prospectus, Warns of Possible Human Extinction

[Financial News]  

AI developer Anthropic finally unveiled its preliminary prospectus (S-1) for an initial public offering (IPO) on the 26th (local time).
According to the Financial Times (FT), the more than 300-page prospectus details ways to cure diseases and grow the economy, on the premise that its technology does not pose “catastrophic or existential risks” to humanity.
Investors in Anthropic, which is set to begin trading on Nasdaq next month under the ticker symbol “ANT,” will have to contend with a complex governance structure that grants ordinary shareholders almost no voting rights, as well as substantial financial and social risks.

Anthropic’s prospectus includes 92 pages of “risk factors,” markedly different from boilerplate disclosures. The section is about three times as long as SpaceX’s filing, the FT reported.
Anthropic warned that its technology could pose “catastrophic or existential risks” to humanity and “could provide meaningful support to malicious actors seeking to develop chemical, biological or other dangerous weapons.”
It also said there was a possibility it could lose control of its AI products. Anthropic warned that AI models might one day exhibit “unexpected ... self-preserving behavior, including resistance to shutdown, information concealment or manipulation,” or generate “coercive and deceptive” outputs resembling “blackmail.” In other words, they could manipulate or conceal information and resist attempts to shut them down.
Anthropic expressed concern that these risks could prove impossible for even a single insurer—or the insurance industry as a whole—to bear.

Anthropic also identified U.S. AI regulatory policy as one factor that could pose a risk.
The administration could raise costs, restrict Anthropic’s operations or reduce demand for its models through new regulations. Earlier this year, Anthropic was designated a “supply chain risk” company by the U.S. Department of Defense. This was because the company had not cooperated with the department’s plan to use AI to attack targets.
Anthropic said in its S-1 that revenue from the government sector accounts for about 1% of its total revenue.

Anthropic noted that supply chain uncertainty is high because the supply of HBM (high-bandwidth memory), a key component of AI accelerators, is concentrated among three companies: Samsung Electronics Co., Ltd., SK hynix and Micron Technology, Inc.
It was particularly concerned that production facilities are located outside the United States and are therefore exposed to export and import controls, sanctions, tariffs and other measures.
Anthropic also warned that AI accelerator manufacturing is heavily concentrated at Taiwan’s TSMC, and that tensions across the Taiwan Strait, military conflict between China and Taiwan, natural disasters and other events could paralyze global semiconductor supply for years.

According to the S-1, Anthropic’s revenue growth has been explosive. Revenue rose from $4.6 billion last year to more than $3.3 billion in the nine months through the end of September this year. The five-year-old company generated annual revenue of $6.6 million per employee last year.
However, substantial risks lurk behind this growth.
In the first nine months of this year, just five customers accounted for 25% of total revenue. Only 250 customers spent $10 million or more annually.
Anthropic was pessimistic that growth could stagnate in the coming months, saying its current rapid growth may “reflect early-stage adoption.”
Revenue growth has slowed as the year has progressed. Revenue more than doubled between the first and second quarters, but the growth rate between the second and third quarters was only around 50%.

Anthropic stressed that AI has enormous potential to transform humanity’s paradigm if the risks of losing control can be overcome.
First, it could have an enormous economic impact, surpassing that of the Industrial Revolution.
Anthropic expects AI to fundamentally transform every sector of the global economy far faster than the Industrial Revolution, the spread of electricity, or the advent of semiconductors and the internet. AI can mass-produce cognitive capabilities themselves—including reasoning, analysis, synthesis, decision-making and creativity—and make them available instantly when needed.
AI is also expected to accelerate disease treatment and scientific discovery.
AI could help humanity find treatments sooner for hard-to-treat diseases such as Alzheimer’s disease and cancer.
AI can complete protein folding in seconds or minutes, a process that previously took anywhere from months to decades. The way a protein folds can completely alter its efficacy, so its three-dimensional structure must be accurately understood to develop new drugs or treatments for hard-to-treat diseases. With AI making it possible to predict that structure almost instantly, the development of new drugs and treatments has accelerated rapidly.
Anthropic was optimistic that, if AI’s potential is properly realized through risk management and safe control, humanity could dramatically improve its quality of life and rapidly accelerate global economic growth.
[email protected] Song Kyung-jae Reporter