When a company plans to sell shares on Wall Street, it must detail the risks faced by investors. In the case of the artificial intelligence company Anthropic, these risks include the possibility that its products threaten nothing less than humanity itself.
These risks appear in the company’s filing with the Securities and Exchange Commission (SEC), the American financial market regulatory authority. They range from the use of AI models by malicious actors to manufacture deadly weapons to the hypothesis of technology escaping all control, with “incalculable” consequences.
Of the 335 pages of the document, devoted notably to finances, executive compensation, and dividend policy, 93 pages are dedicated to “risk factors,” that is, what could harm potential shareholders’ investments. Its competitor SpaceX counts 38 pages on over 250 pages. To attract investors, Anthropic presents its IPO as a step toward reducing the risks that, according to it, AI poses.
In a letter addressed to future shareholders, the founders write: “Building AI is about harnessing cognition itself, and anything so powerful can be used for both good and evil.” They add that Anthropic is “an organization deeply guided by its values, largely founded to help the world harness the benefits of AI and avoid its dangers.”
The risk of out-of-control technology
Anthropic indicates that its models could develop so rapidly that the company fails to implement adequate safety measures. Such a situation “could increase the likelihood of security incidents, regulatory interventions, or reputational harm, and negatively affect its business.” CEO Dario Amodei warned in September that within six to twelve months, AI could be capable of directing a swarm of agents able to take control of the entire internet.
According to the filing, as systems become “more autonomous and capable of making consequential decisions with less human supervision,” they could independently take harmful actions on a large scale. The company specifies that even if its own models do not cause such damage, a serious security incident involving cutting-edge AI could harm the entire sector.
Malicious uses and computational limits
The document also mentions the possibility that advanced systems, if not equipped with sufficient safeguards, “pose catastrophic or existential risks to humanity.” Anthropic notably cites the manufacture of chemical weapons and the use of technology by authoritarian leaders to repress their citizens. In September, the company indicated it had blocked attempts to use its models for malicious purposes, including cyberattacks, surveillance, and research potentially leading to the development of biological weapons.
The company also mentions computational resources, that is, the computing power necessary to train, test, and improve AI models, which are costly and in high demand. The filing further highlights that building data centers has sparked “public controversies and regulatory scrutiny” regarding their environmental impact, land and water use, and allocation of limited energy resources.
A strategy based on safety
Anthropic has positioned itself as a responsible AI company and emphasizes its vigilance regarding safety to its professional clients. “We believe that cutting-edge AI companies must invest in safety and alignment,” the filing states, to ensure responsible deployment of this technology.
Brian Mulberry, chief market strategist at Zacks Investment Research, emphasizes that IPOs must present the risks faced by investors “very clearly.” According to him, this is one of the strengths of American capital markets: being listed on a U.S. exchange requires exposing these risks “very clearly,” so that everyone understands what they are financing by investing their money.
