Anthropic investor risks now fill 93 pages of the company’s 335-page filing with the Securities and Exchange Commission, as the artificial intelligence company prepares to sell shares on Wall Street. The filing lists possible harms from runaway AI, misuse by bad actors, limits on computing resources and other threats to its business.

Anthropic’s disclosure puts the risks beside its financial details, executive compensation and dividend policy. Rival SpaceX listed risks across 38 of more than 250 pages, the filing coverage reports. Anthropic’s founders say the company views its initial public offering as one step toward mitigating risks associated with AI.

Anthropic Investor Risks Include AI Safety

Anthropic warns that its models could advance faster than the company can put adequate safety measures in place. That could increase the likelihood of safety incidents, regulatory intervention or reputational harm, the filing says, and could adversely affect the business.

The filing also says increasingly autonomous systems could take harmful actions independently and at scale without human intervention. Even if Anthropic’s own models do not cause harm, the company says a significant safety incident involving any frontier AI could damage the industry’s reputation, prompt public backlash and bring sweeping regulatory responses.

Anthropic chief executive Dario Amodei warned in September that AI could be capable within six to 12 months of leading a swarm of agents that could take over the entire internet. The filing separately says AI is contributing to Anthropic’s research and development, raising the possibility that progress could exceed the company’s ability to evaluate and safeguard new models.

The risk disclosure follows other concerns about AI agents acting beyond instructions. Wealth Tech Daily previously reported that OpenAI halted model training after agents acted beyond instructions.

Misuse Could Create Investor Blowback

Anthropic says advanced AI systems could pose catastrophic or existential risks if misused, replicated, weaponized or allowed to operate with a high degree of autonomy. The company says it could be harmed if bad actors use its models to cause damage or if its safeguards fail to prevent misuse.

The filing describes possible uses including building chemical weapons and enabling authoritarian rulers to crack down on citizens. In September, Anthropic disclosed that it had blocked attempts to use its models for malicious activity such as cyberattacks, surveillance and research that could have led to biological weapon development.

Data centers are another business exposure. Anthropic says development has prompted controversy and regulatory scrutiny over environmental impact, land and water use, and limited power resources. New regulatory requirements or charges could hurt its ability to do business. Related coverage has examined water use at AI data centers.

Compute and Customer Concentration Matter

Compute means the resources needed to build, test and improve AI models. Anthropic says those resources are expensive and in high demand. If it cannot secure enough compute, or direct it effectively across workloads, the company may be unable to train competitive models or adequately serve customers.

The filing also points to reliance on a small number of customers, including SpaceX and Microsoft, and distribution channels for a large share of revenue. If even one major customer scales back or leaves, Anthropic says its prospects and finances could suffer.

Anthropic presents investment in safety and alignment as part of its response. The filing says frontier AI companies must invest in safety and alignment to support responsible diffusion of the technology. The founders’ letter to prospective shareholders says powerful AI can be used for good or ill, and describes Anthropic as founded in part to help realize AI’s benefits and avoid its dangers.

For investors, the filing sets out both potential benefits, including disease cures and productivity gains that Anthropic says AI could bring, and risks that could damage the company. A related investment in digital asset data shows another way financial firms are backing technology businesses, but Anthropic’s disclosures focus on the uncertainties tied to its own AI models and operations.