Risks Associated with AI Highlighted in IPO Filing by Anthropic

A recent IPO filing by Anthropic highlights a stark imbalance in how risks associated with artificial intelligence (AI) are presented to investors and the general public. The 240,000-word document devotes an extraordinary amount of space, approximately 69,800 words or 30%, to detailing potential pitfalls for shareholders.

These risks include catastrophic events, job displacement, rogue AI agents, industry concentration, and even litigation stemming from Anthropic's own statements on AI hazards. This extensive coverage of potential costs to investors is a testament to the traditional purpose of an IPO prospectus, warning those providing capital about what might go wrong.

However, there is another crucial aspect that such filings often overlook: the impact on society as a whole. In contrast to its detailed breakdown of risks for shareholders, Anthropic's prospectus remains largely silent on the potential costs and consequences of its AI products for the broader public.

This disparity raises questions about accountability and transparency in the development and deployment of advanced technologies like AI. While investors are provided with extensive warnings, those who will bear the actual costs and consequences, often without their consent or direct benefit, are not afforded similar consideration.

A comprehensive Register of AI Risks & Remedies has been compiled through research collaborations, including one with Oxfam. This living document catalogues 103 current AI risks, prioritizing them by severity, identifying who bears the costs, and assessing the readiness of remedies.

This register provides a stark contrast to Anthropic's prospectus, framing the risks as they apply not just to investors but also to society at large. It serves as a reminder that while investors may be warned about potential losses, the true impact of AI technologies on humanity is often left unaddressed and unseen in such filings.

The Delaware Public Benefit Corporation status of Anthropic raises questions about its accountability to those affected by its actions. As a corporation subject to Delaware law, Anthropic is required to report on how it has served "those materially affected by the corporation's conduct". However, its disclosures in this area focus on efforts rather than tangible results.

The company's charter emphasizes the need to balance financial interests with the benefits of the broader public. To fulfill this obligation, a separate document detailing the public risks associated with Anthropic's AI technologies could be beneficial. This "Public Risk Statement" would provide a comprehensive overview of the potential consequences for individuals and communities worldwide.

Anthropic's existing prospectus devotes 69,800 words to discussing the risks associated with its AI technologies for investors. However, this focus on investor risk overlooks the broader implications of these technologies. A review of publicly available records reveals numerous instances where AI has had devastating effects on individuals and communities.

In some cases, people have been mistakenly flagged by fraud algorithms, leading to personal and financial consequences. For example, a person in the Netherlands may be incorrectly identified as a potential threat, while a Polish doctor may see their skills eroded due to over-reliance on an algorithm. In other instances, individuals like Kenyan content moderators are tasked with labeling toxic content, resulting in traumatic experiences.

Three key considerations should be included in any Public Risk Statement. Firstly, the impact of AI on workers and worker productivity is a pressing concern. Research has shown that overreliance on AI can lead to deskilling, as people become less proficient in their roles due to automation bias.

Furthermore, human oversight can fail under pressure to improve productivity, as acknowledged by Anthropic's prospectus. The company's mention of overreliance in passing and complete absence of deskilling discussion raises questions about its commitment to transparency on this issue.

The mention of oversight in Anthropic's IPO filing raises important questions about who is responsible for reviewing AI decisions and whether they have the necessary authority to correct them if needed.

Investors are being told that the company's productivity benefits will lead to economic gains, but this assumes that the costs associated with checking the work are negligible. However, when these costs are factored in, the actual gains may be significantly lower than anticipated.

The prospectus highlights several instances where AI has led to costly mistakes for individuals and organizations, including wrongful arrests, discriminatory screening, and copyright infringement. These cases demonstrate how widespread use of AI can expose more people to errors and potential financial losses.

As these precedents continue to build, the demand for compensation may rise, potentially eroding the economic gains that Anthropic is counting on. This could have significant implications for the company's sales numbers and overall revenue.

Anthropic's own experience with a $1.5 billion settlement over pirated books serves as a reminder of the potential risks associated with AI use.

The complexities surrounding artificial intelligence's impact on society are becoming increasingly apparent, and investors are taking notice.

A closer examination of AI-related settlements reveals a surprising trend: most were resolved using existing laws, rather than requiring new legislation. This raises questions about the relationship between legal protection and practical means for securing redress.

As governments grapple with the challenges posed by AI, state legislatures are stepping in to fill the void. In 2026, Transparency Coalition reported that 30 states had enacted 115 AI-related laws, a nearly 58% increase from the previous year. This trend suggests that even if federal action is slow or non-existent, individual states and regulatory bodies will continue to address AI concerns.

State attorneys-general are also taking action against companies they believe have harmed children through their AI products. Eight AGs have filed lawsuits against OpenAI, while the FTC is investigating safety lapses at Frontier Labs. Meanwhile, the EU is launching enforcement efforts under its AI Act, and the Financial Stability Board is urging banks to strengthen their AI governance.

These developments signal that AI-related losses will become a cost of doing business as companies are increasingly held accountable for their products' impacts. As investors scrutinize Anthropic's IPO filing, they may demand greater transparency about the company's reliance on external costs and subsidies.

In light of these emerging trends, it is reasonable to expect that companies like Anthropic will soon be required to disclose more information about their AI products' potential risks and liabilities.

As companies like Anthropic continue to develop advanced AI technologies, there is growing concern about the potential risks these products pose to society as a whole.

Regulatory bodies may soon demand that companies disclose more information about their AI products' potential risks and liabilities, in addition to those affecting investors. This shift in focus could lead to greater transparency and accountability in the development of AI technologies. Ultimately, this move towards increased disclosure is likely to have far-reaching implications for the industry as a whole.