This institution has been granted access to the Anthropic Investor prospectus Reuters continues to squeeze out bits and pieces of information about the risk factors the startup considers critical to its own business. In addition to the potential for unscrupulous AI agents to expose the company to lawsuits, Anthropic cited its reliance on infrastructure from large third-party companies as a risk factor.
Image source: Human
“These autonomous capabilities increase the potential for corruption, errors, breaches, and exploits of security vulnerabilities that can have real-world consequences.” — Anthropic describes the threat posed by rogue artificial intelligence agents in its prospectus. The limits of liability set out in Anthropic’s contracts with users of its artificial intelligence systems may not be sufficient to cover possible damages and claims by persons affected by the actions of these customers.
Representatives of the startup said that existing legal norms do not regulate the active use of artificial intelligence agents, so the legal consequences of such incidents remain difficult to predict. From a legal perspective, the behavior of artificial intelligence agents is difficult to explain using traditional practices. First, the status of agents themselves within the legal realm is unclear. Second, the degree of liability of direct users of these AI agents to third parties has not yet been determined. Anthropic acknowledges that its security measures may be insufficient to prevent harm, including to the Company itself.
Anthropic also acknowledged in its investor prospectus that its business is dependent on a small number of partners and customers and that such dependence poses significant risks. Last year, up to 47% of Anthropic’s services were sold through Amazon (AWS) and Alphabet (Google) cloud infrastructure. Both companies are major customers of Anthropic and compete with it in the AI services market. Under adverse circumstances, this relationship may worsen the competitive position of the new venture. In 2023, these companies accounted for only 11% of startup revenue, but by 2024, their share had increased to 32%.
Anthropic’s revenue grew 12 times last year to $4.6 billion, mainly from Claude customers ($3.8 billion) and subscribers ($789 million). As the startup points out, variable revenue determined by fees for using Cloud will continue to dominate its structure for the foreseeable future. Anthropic generated 47% of its total revenue, or $2.16 billion, through the cloud market last year. At the same time, it also paid $351 million in commissions to the platform. In fact, cloud providers earn an average of 16 cents for every dollar Anthropic makes.
Anthropic’s agreements with suppliers and data center builders totaled $54.6 billion by the end of 2025, but by the beginning of this year, long-term agreements in this area had reached $417 billion, equivalent to 3.5 GW of computing power. The company sees partnerships with Amazon, Google and Microsoft (Azure) in this space as an advantage as it can increase sales of its services faster and reach more customers. At the same time, as Anthropic acknowledges, relying on only a few large market players carries certain risks. In addition to this, the company’s partners can obtain information about its pricing and marketing policies.
By the end of 2025, Anthropic partners will be responsible for 60% of invoices submitted to startup customers, totaling $909 million, while by 2024, this share will not exceed 42%. Even if Anthropic has direct agreements with customers, the cloud giant’s role as a payment agent may pose certain risks to the startup’s revenue growth rate. In addition, Anthropic’s largest customers accounted for 12% of the company’s total revenue last year, and this concentration also brings certain risks to the business.
If you find an error, select it with your mouse and press CTRL+ENTER.
